Daily Papers

Last year, multimodal architectures served up a revolution in AI-based approaches and solutions, extending the capabilities of large language models (LLM). We propose an OmniFusion model based on a pretrained LLM and adapters for visual modality. We evaluated and compared several architecture design principles for better text and visual data coupling: MLP and transformer adapters, various CLIP ViT-based encoders (SigLIP, InternVIT, etc.), and their fusing approach, image encoding method (whole image or tiles encoding) and two 7B LLMs (the proprietary one and open-source Mistral). Experiments on 8 visual-language benchmarks show the top score for the best OmniFusion setup in terms of different VQA tasks in comparison with open-source LLaVA-like solutions: VizWiz, Pope, MM-Vet, ScienceQA, MMBench, TextVQA, VQAv2, MMMU. We also propose a variety of situations, where OmniFusion provides highly-detailed answers in different domains: housekeeping, sightseeing, culture, medicine, handwritten and scanned equations recognition, etc. Mistral-based OmniFusion model is an open-source solution with weights, training and inference scripts available at https://github.com/AIRI-Institute/OmniFusion.

Recent advancements in Multimodal Large Language Models (MLLMs) have revolutionized the field of vision-language understanding by integrating visual perception capabilities into Large Language Models (LLMs). The prevailing trend in this field involves the utilization of a vision encoder derived from vision-language contrastive learning (CL), showing expertise in capturing overall representations while facing difficulties in capturing detailed local patterns. In this work, we focus on enhancing the visual representations for MLLMs by combining high-frequency and detailed visual representations, obtained through masked image modeling (MIM), with semantically-enriched low-frequency representations captured by CL. To achieve this goal, we introduce X-Former which is a lightweight transformer module designed to exploit the complementary strengths of CL and MIM through an innovative interaction mechanism. Specifically, X-Former first bootstraps vision-language representation learning and multimodal-to-multimodal generative learning from two frozen vision encoders, i.e., CLIP-ViT (CL-based) and MAE-ViT (MIM-based). It further bootstraps vision-to-language generative learning from a frozen LLM to ensure visual features from X-Former can be interpreted by the LLM. To demonstrate the effectiveness of our approach, we assess its performance on tasks demanding detailed visual understanding. Extensive evaluations indicate that X-Former excels in visual reasoning tasks involving both structural and semantic categories in the GQA dataset. Assessment on fine-grained visual perception benchmark further confirms its superior capabilities in visual understanding.

Pioneering token-based works such as Chameleon and Emu3 have established a foundation for multimodal unification but face challenges of high training computational overhead and limited comprehension performance due to a lack of high-level semantics. In this paper, we introduce TokLIP, a visual tokenizer that enhances comprehension by semanticizing vector-quantized (VQ) tokens and incorporating CLIP-level semantics while enabling end-to-end multimodal autoregressive training with standard VQ tokens. TokLIP integrates a low-level discrete VQ tokenizer with a ViT-based token encoder to capture high-level continuous semantics. Unlike previous approaches (e.g., VILA-U) that discretize high-level features, TokLIP disentangles training objectives for comprehension and generation, allowing the direct application of advanced VQ tokenizers without the need for tailored quantization operations. Our empirical results demonstrate that TokLIP achieves exceptional data efficiency, empowering visual tokens with high-level semantic understanding while enhancing low-level generative capacity, making it well-suited for autoregressive Transformers in both comprehension and generation tasks. The code and models are available at https://github.com/TencentARC/TokLIP.

Large-scale vision-language pre-training has achieved promising results on downstream tasks. Existing methods highly rely on the assumption that the image-text pairs crawled from the Internet are in perfect one-to-one correspondence. However, in real scenarios, this assumption can be difficult to hold: the text description, obtained by crawling the affiliated metadata of the image, often suffers from the semantic mismatch and the mutual compatibility. To address these issues, we introduce PyramidCLIP, which constructs an input pyramid with different semantic levels for each modality, and aligns visual elements and linguistic elements in the form of hierarchy via peer-level semantics alignment and cross-level relation alignment. Furthermore, we soften the loss of negative samples (unpaired samples) so as to weaken the strict constraint during the pre-training stage, thus mitigating the risk of forcing the model to distinguish compatible negative pairs. Experiments on five downstream tasks demonstrate the effectiveness of the proposed PyramidCLIP. In particular, with the same amount of 15 million pre-training image-text pairs, PyramidCLIP exceeds CLIP on ImageNet zero-shot classification top-1 accuracy by 10.6%/13.2%/10.0% with ResNet50/ViT-B32/ViT-B16 based image encoder respectively. When scaling to larger datasets, PyramidCLIP achieves the state-of-the-art results on several downstream tasks. In particular, the results of PyramidCLIP-ResNet50 trained on 143M image-text pairs surpass that of CLIP using 400M data on ImageNet zero-shot classification task, significantly improving the data efficiency of CLIP.

The architecture of multimodal large language models (MLLMs) commonly connects a vision encoder, often based on CLIP-ViT, to a large language model. While CLIP-ViT works well for capturing global image features, it struggles to model local relationships between adjacent patches, leading to weaker visual representation, which in turn affects the detailed understanding ability of MLLMs. To solve this, we propose LLaVA-SP, which only adds six spatial visual tokens to the original visual tokens to enhance the visual representation. Our approach offers three key advantages: 1)We propose a novel Projector, which uses convolutional kernels to derive visual spatial tokens from ViT patch features, simulating two visual spatial ordering approaches: ``from central region to global" and ``from abstract to specific". Then, a cross-attention mechanism is applied to fuse fine-grained visual information, enriching the overall visual representation. 2) We present two model variants: LLaVA-SP-Cropping, which focuses on detail features through progressive cropping, and LLaVA-SP-Pooling, which captures global semantics through adaptive pooling, enabling the model to handle diverse visual understanding tasks. 3) Extensive experiments show that LLaVA-SP, fine-tuned with LoRA, achieves significant performance improvements across various multimodal benchmarks, outperforming the state-of-the-art LLaVA-1.5 model in multiple tasks with nearly identical inference latency. The code and models are available at https://github.com/CnFaker/LLaVA-SP.

In this paper, we tackle an emerging computer vision task, open-vocabulary universal image segmentation, that aims to perform semantic/instance/panoptic segmentation (background semantic labeling + foreground instance segmentation) for arbitrary categories of text-based descriptions in inference time. We first build a baseline method by directly adopting pre-trained CLIP models without finetuning or distillation. We then develop MaskCLIP, a Transformer-based approach with a MaskCLIP Visual Encoder, which is an encoder-only module that seamlessly integrates mask tokens with a pre-trained ViT CLIP model for semantic/instance segmentation and class prediction. MaskCLIP learns to efficiently and effectively utilize pre-trained partial/dense CLIP features within the MaskCLIP Visual Encoder that avoids the time-consuming student-teacher training process. MaskCLIP outperforms previous methods for semantic/instance/panoptic segmentation on ADE20K and PASCAL datasets. We show qualitative illustrations for MaskCLIP with online custom categories. Project website: https://maskclip.github.io.

We present a simple approach which can turn a ViT encoder into an efficient video model, which can seamlessly work with both image and video inputs. By sparsely sampling the inputs, the model is able to do training and inference from both inputs. The model is easily scalable and can be adapted to large-scale pre-trained ViTs without requiring full finetuning. The model achieves SOTA results and the code will be open-sourced.

In this paper, we present a light-weight detection transformer, LW-DETR, which outperforms YOLOs for real-time object detection. The architecture is a simple stack of a ViT encoder, a projector, and a shallow DETR decoder. Our approach leverages recent advanced techniques, such as training-effective techniques, e.g., improved loss and pretraining, and interleaved window and global attentions for reducing the ViT encoder complexity. We improve the ViT encoder by aggregating multi-level feature maps, and the intermediate and final feature maps in the ViT encoder, forming richer feature maps, and introduce window-major feature map organization for improving the efficiency of interleaved attention computation. Experimental results demonstrate that the proposed approach is superior over existing real-time detectors, e.g., YOLO and its variants, on COCO and other benchmark datasets. Code and models are available at (https://github.com/Atten4Vis/LW-DETR).

CLIP models perform remarkably well on zero-shot classification and retrieval tasks. But recent studies have shown that learnt representations in CLIP are not well suited for dense prediction tasks like object detection, semantic segmentation or depth estimation. More recently, multi-stage training methods for CLIP models was introduced to mitigate the weak performance of CLIP on downstream tasks. In this work, we find that simply improving the quality of captions in image-text datasets improves the quality of CLIP's visual representations, resulting in significant improvement on downstream dense prediction vision tasks. In fact, we find that CLIP pretraining with good quality captions can surpass recent supervised, self-supervised and weakly supervised pretraining methods. We show that when CLIP model with ViT-B/16 as image encoder is trained on well aligned image-text pairs it obtains 12.1% higher mIoU and 11.5% lower RMSE on semantic segmentation and depth estimation tasks over recent state-of-the-art Masked Image Modeling (MIM) pretraining methods like Masked Autoencoder (MAE). We find that mobile architectures also benefit significantly from CLIP pretraining. A recent mobile vision architecture, MCi2, with CLIP pretraining obtains similar performance as Swin-L, pretrained on ImageNet-22k for semantic segmentation task while being 6.1times smaller. Moreover, we show that improving caption quality results in 10times data efficiency when finetuning for dense prediction tasks.

Recent studies have shown that CLIP has achieved remarkable success in performing zero-shot inference while its fine-tuning performance is not satisfactory. In this paper, we identify that fine-tuning performance is significantly impacted by hyper-parameter choices. We examine various key hyper-parameters and empirically evaluate their impact in fine-tuning CLIP for classification tasks through a comprehensive study. We find that the fine-tuning performance of CLIP is substantially underestimated. Equipped with hyper-parameter refinement, we demonstrate CLIP itself is better or at least competitive in fine-tuning compared with large-scale supervised pre-training approaches or latest works that use CLIP as prediction targets in Masked Image Modeling. Specifically, CLIP ViT-Base/16 and CLIP ViT-Large/14 can achieve 85.7%,88.0% finetuning Top-1 accuracy on the ImageNet-1K dataset . These observations challenge the conventional conclusion that CLIP is not suitable for fine-tuning, and motivate us to rethink recently proposed improvements based on CLIP. We will release our code publicly at https://github.com/LightDXY/FT-CLIP.

Contrastive pre-training on image-text pairs, exemplified by CLIP, becomes a standard technique for learning multi-modal visual-language representations. Although CLIP has demonstrated remarkable performance, training it from scratch on noisy web-scale datasets is computationally demanding. On the other hand, mask-then-predict pre-training approaches, like Masked Image Modeling (MIM), offer efficient self-supervised learning for single-modal representations. This paper introduces Unmasked Token Alignment (UTA), a method that leverages existing CLIP models to further enhance its vision-language representations. UTA trains a Vision Transformer (ViT) by aligning unmasked visual tokens to the corresponding image tokens from a frozen CLIP vision encoder, which automatically aligns the ViT model with the CLIP text encoder. The pre-trained ViT can be directly applied for zero-shot evaluation even without training on image-text pairs. Compared to MIM approaches, UTA does not suffer from training-finetuning inconsistency and is much more training-efficient by avoiding using the extra [MASK] tokens. Extensive experimental results demonstrate that UTA can enhance CLIP models and outperform existing MIM methods on various uni- and multi-modal benchmarks. Code and models are available at https://github.com/jihaonew/UTA.

Contrastive Language-Image Pre-training (CLIP) excels in global alignment with language but exhibits limited sensitivity to spatial information, leading to strong performance in zero-shot classification tasks but underperformance in tasks requiring precise spatial understanding. Recent approaches have introduced Region-Language Alignment (RLA) to enhance CLIP's performance in dense multimodal tasks by aligning regional visual representations with corresponding text inputs. However, we find that CLIP ViTs fine-tuned with RLA suffer from notable loss in spatial awareness, which is crucial for dense prediction tasks. To address this, we propose the Spatial Correlation Distillation (SCD) framework, which preserves CLIP's inherent spatial structure and mitigates the above degradation. To further enhance spatial correlations, we introduce a lightweight Refiner that extracts refined correlations directly from CLIP before feeding them into SCD, based on an intriguing finding that CLIP naturally captures high-quality dense features. Together, these components form a robust distillation framework that enables CLIP ViTs to integrate both visual-language and visual-centric improvements, achieving state-of-the-art results across various open-vocabulary dense prediction benchmarks.

The feature maps of vision encoders are fundamental to myriad modern AI tasks, ranging from core perception algorithms (e.g. semantic segmentation, object detection, depth perception, etc.) to modern multimodal understanding in vision-language models (VLMs). Currently, in computer vision, the frontier of general purpose vision backbones is Vision Transformers (ViT), typically trained using contrastive loss (e.g. CLIP). A key problem with most off-the-shelf ViTs, particularly CLIP, is that these models are inflexibly low resolution. Most run at 224 times 224px, while the "high-resolution" versions are around 378-448px, but still inflexible. We introduce a novel method to coherently and cheaply upsample the feature maps of low-resolution vision encoders while picking up on fine-grained details that would otherwise be lost due to resolution. We demonstrate the effectiveness of this approach on core perception tasks as well as within agglomerative model training using RADIO as a way of providing richer targets for distillation. Code available at https://github.com/NVlabs/FeatSharp .

Open-vocabulary dense prediction tasks including object detection and image segmentation have been advanced by the success of Contrastive Language-Image Pre-training (CLIP). CLIP models, particularly those incorporating vision transformers (ViTs), have exhibited remarkable generalization ability in zero-shot image classification. However, when transferring the vision-language alignment of CLIP from global image representation to local region representation for the open-vocabulary dense prediction tasks, CLIP ViTs suffer from the domain shift from full images to local image regions. In this paper, we embark on an in-depth analysis of the region-language alignment in CLIP models, which is essential for downstream open-vocabulary dense prediction tasks. Subsequently, we propose an approach named CLIPSelf, which adapts the image-level recognition ability of CLIP ViT to local image regions without needing any region-text pairs. CLIPSelf empowers ViTs to distill itself by aligning a region representation extracted from its dense feature map with the image-level representation of the corresponding image crop. With the enhanced CLIP ViTs, we achieve new state-of-the-art performance on open-vocabulary object detection, semantic segmentation, and panoptic segmentation across various benchmarks. Models and code will be available at https://github.com/wusize/CLIPSelf.

Vision transformers (ViTs) have gained popularity recently. Even without customized image operators such as convolutions, ViTs can yield competitive performance when properly trained on massive data. However, the computational overhead of ViTs remains prohibitive, due to stacking multi-head self-attention modules and else. Compared to the vast literature and prevailing success in compressing convolutional neural networks, the study of Vision Transformer compression has also just emerged, and existing works focused on one or two aspects of compression. This paper proposes a unified ViT compression framework that seamlessly assembles three effective techniques: pruning, layer skipping, and knowledge distillation. We formulate a budget-constrained, end-to-end optimization framework, targeting jointly learning model weights, layer-wise pruning ratios/masks, and skip configurations, under a distillation loss. The optimization problem is then solved using the primal-dual algorithm. Experiments are conducted with several ViT variants, e.g. DeiT and T2T-ViT backbones on the ImageNet dataset, and our approach consistently outperforms recent competitors. For example, DeiT-Tiny can be trimmed down to 50\% of the original FLOPs almost without losing accuracy. Codes are available online:~https://github.com/VITA-Group/UVC.

Recent breakthroughs in vision-language models (VLMs) start a new page in the vision community. The VLMs provide stronger and more generalizable feature embeddings compared to those from ImageNet-pretrained models, thanks to the training on the large-scale Internet image-text pairs. However, despite the amazing achievement from the VLMs, vanilla Vision Transformers (ViTs) remain the default choice for the image encoder. Although pure transformer proves its effectiveness in the text encoding area, it remains questionable whether it is also the case for image encoding, especially considering that various types of networks are proposed on the ImageNet benchmark, which, unfortunately, are rarely studied in VLMs. Due to small data/model scale, the original conclusions of model design on ImageNet can be limited and biased. In this paper, we aim at building an evaluation protocol of vision models in the vision-language era under the contrastive language-image pretraining (CLIP) framework. We provide a comprehensive way to benchmark different vision models, covering their zero-shot performance and scalability in both model and training data sizes. To this end, we introduce ViTamin, a new vision models tailored for VLMs. ViTamin-L significantly outperforms ViT-L by 2.0% ImageNet zero-shot accuracy, when using the same publicly available DataComp-1B dataset and the same OpenCLIP training scheme. ViTamin-L presents promising results on 60 diverse benchmarks, including classification, retrieval, open-vocabulary detection and segmentation, and large multi-modal models. When further scaling up the model size, our ViTamin-XL with only 436M parameters attains 82.9% ImageNet zero-shot accuracy, surpassing 82.0% achieved by EVA-E that has ten times more parameters (4.4B).

This paper investigates the performance of the Contrastive Language-Image Pre-training (CLIP) when scaled down to limited computation budgets. We explore CLIP along three dimensions: data, architecture, and training strategies. With regards to data, we demonstrate the significance of high-quality training data and show that a smaller dataset of high-quality data can outperform a larger dataset with lower quality. We also examine how model performance varies with different dataset sizes, suggesting that smaller ViT models are better suited for smaller datasets, while larger models perform better on larger datasets with fixed compute. Additionally, we provide guidance on when to choose a CNN-based architecture or a ViT-based architecture for CLIP training. We compare four CLIP training strategies - SLIP, FLIP, CLIP, and CLIP+Data Augmentation - and show that the choice of training strategy depends on the available compute resource. Our analysis reveals that CLIP+Data Augmentation can achieve comparable performance to CLIP using only half of the training data. This work provides practical insights into how to effectively train and deploy CLIP models, making them more accessible and affordable for practical use in various applications.

Large multimodal datasets have been instrumental in recent breakthroughs such as CLIP, Stable Diffusion, and GPT-4. At the same time, datasets rarely receive the same research attention as model architectures or training algorithms. To address this shortcoming in the machine learning ecosystem, we introduce DataComp, a benchmark where the training code is fixed and researchers innovate by proposing new training sets. We provide a testbed for dataset experiments centered around a new candidate pool of 12.8B image-text pairs from Common Crawl. Participants in our benchmark design new filtering techniques or curate new data sources and then evaluate their new dataset by running our standardized CLIP training code and testing on 38 downstream test sets. Our benchmark consists of multiple scales, with four candidate pool sizes and associated compute budgets ranging from 12.8M to 12.8B samples seen during training. This multi-scale design facilitates the study of scaling trends and makes the benchmark accessible to researchers with varying resources. Our baseline experiments show that the DataComp workflow is a promising way of improving multimodal datasets. We introduce DataComp-1B, a dataset created by applying a simple filtering algorithm to the 12.8B candidate pool. The resulting 1.4B subset enables training a CLIP ViT-L/14 from scratch to 79.2% zero-shot accuracy on ImageNet. Our new ViT-L/14 model outperforms a larger ViT-g/14 trained on LAION-2B by 0.7 percentage points while requiring 9x less training compute. We also outperform OpenAI's CLIP ViT-L/14 by 3.7 percentage points, which is trained with the same compute budget as our model. These gains highlight the potential for improving model performance by carefully curating training sets. We view DataComp-1B as only the first step and hope that DataComp paves the way toward the next generation of multimodal datasets.

We delve into a nuanced but significant challenge inherent to Vision Transformers (ViTs): feature maps of these models exhibit grid-like artifacts, which detrimentally hurt the performance of ViTs in downstream tasks. Our investigations trace this fundamental issue down to the positional embeddings at the input stage. To address this, we propose a novel noise model, which is universally applicable to all ViTs. Specifically, the noise model dissects ViT outputs into three components: a semantics term free from noise artifacts and two artifact-related terms that are conditioned on pixel locations. Such a decomposition is achieved by enforcing cross-view feature consistency with neural fields in a per-image basis. This per-image optimization process extracts artifact-free features from raw ViT outputs, providing clean features for offline applications. Expanding the scope of our solution to support online functionality, we introduce a learnable denoiser to predict artifact-free features directly from unprocessed ViT outputs, which shows remarkable generalization capabilities to novel data without the need for per-image optimization. Our two-stage approach, termed Denoising Vision Transformers (DVT), does not require re-training existing pre-trained ViTs and is immediately applicable to any Transformer-based architecture. We evaluate our method on a variety of representative ViTs (DINO, MAE, DeiT-III, EVA02, CLIP, DINOv2, DINOv2-reg). Extensive evaluations demonstrate that our DVT consistently and significantly improves existing state-of-the-art general-purpose models in semantic and geometric tasks across multiple datasets (e.g., +3.84 mIoU). We hope our study will encourage a re-evaluation of ViT design, especially regarding the naive use of positional embeddings.

Vision transformers (ViTs) have demonstrated remarkable performance across various visual tasks. However, ViT models suffer from substantial computational and memory requirements, making it challenging to deploy them on resource-constrained platforms. Quantization is a popular approach for reducing model size, but most studies mainly focus on equal bit-width quantization for the entire network, resulting in sub-optimal solutions. While there are few works on mixed precision quantization (MPQ) for ViTs, they typically rely on search space-based methods or employ mixed precision arbitrarily. In this paper, we introduce LRP-QViT, an explainability-based method for assigning mixed-precision bit allocations to different layers based on their importance during classification. Specifically, to measure the contribution score of each layer in predicting the target class, we employ the Layer-wise Relevance Propagation (LRP) method. LRP assigns local relevance at the output layer and propagates it through all layers, distributing the relevance until it reaches the input layers. These relevance scores serve as indicators for computing the layer contribution score. Additionally, we have introduced a clipped channel-wise quantization aimed at eliminating outliers from post-LayerNorm activations to alleviate severe inter-channel variations. To validate and assess our approach, we employ LRP-QViT across ViT, DeiT, and Swin transformer models on various datasets. Our experimental findings demonstrate that both our fixed-bit and mixed-bit post-training quantization methods surpass existing models in the context of 4-bit and 6-bit quantization.

Contrastive pretraining of image-text foundation models, such as CLIP, demonstrated excellent zero-shot performance and improved robustness on a wide range of downstream tasks. However, these models utilize large transformer-based encoders with significant memory and latency overhead which pose challenges for deployment on mobile devices. In this work, we introduce MobileCLIP -- a new family of efficient image-text models optimized for runtime performance along with a novel and efficient training approach, namely multi-modal reinforced training. The proposed training approach leverages knowledge transfer from an image captioning model and an ensemble of strong CLIP encoders to improve the accuracy of efficient models. Our approach avoids train-time compute overhead by storing the additional knowledge in a reinforced dataset. MobileCLIP sets a new state-of-the-art latency-accuracy tradeoff for zero-shot classification and retrieval tasks on several datasets. Our MobileCLIP-S2 variant is 2.3times faster while more accurate compared to previous best CLIP model based on ViT-B/16. We further demonstrate the effectiveness of our multi-modal reinforced training by training a CLIP model based on ViT-B/16 image backbone and achieving +2.9% average performance improvement on 38 evaluation benchmarks compared to the previous best. Moreover, we show that the proposed approach achieves 10times-1000times improved learning efficiency when compared with non-reinforced CLIP training.

Though the success of CLIP-based training recipes in vision-language models, their scalability to more modalities (e.g., 3D, audio, etc.) is limited to large-scale data, which is expensive or even inapplicable for rare modalities. In this paper, we present ViT-Lens that facilitates efficient omni-modal representation learning by perceiving novel modalities with a pretrained ViT and aligning to a pre-defined space. Specifically, the modality-specific lens is tuned to project multimodal signals to the shared embedding space, which are then processed by a strong ViT that carries pre-trained image knowledge. The encoded multimodal representations are optimized toward aligning with the modal-independent space, pre-defined by off-the-shelf foundation models. A well-trained lens with a ViT backbone has the potential to serve as one of these foundation models, supervising the learning of subsequent modalities. ViT-Lens provides a unified solution for representation learning of increasing modalities with two appealing benefits: (i) Exploiting the pretrained ViT across tasks and domains effectively with efficient data regime; (ii) Emergent downstream capabilities of novel modalities are demonstrated due to the modality alignment space. We evaluate ViT-Lens in the context of 3D as an initial verification. In zero-shot 3D classification, ViT-Lens achieves substantial improvements over previous state-of-the-art, showing 52.0% accuracy on Objaverse-LVIS, 87.4% on ModelNet40, and 60.6% on ScanObjectNN. Furthermore, we enable zero-shot 3D question-answering by simply integrating the trained 3D lens into the InstructBLIP model without any adaptation. We will release the results of ViT-Lens on more modalities in the near future.

Most existing Vision-and-Language (V&L) models rely on pre-trained visual encoders, using a relatively small set of manually-annotated data (as compared to web-crawled data), to perceive the visual world. However, it has been observed that large-scale pretraining usually can result in better generalization performance, e.g., CLIP (Contrastive Language-Image Pre-training), trained on a massive amount of image-caption pairs, has shown a strong zero-shot capability on various vision tasks. To further study the advantage brought by CLIP, we propose to use CLIP as the visual encoder in various V&L models in two typical scenarios: 1) plugging CLIP into task-specific fine-tuning; 2) combining CLIP with V&L pre-training and transferring to downstream tasks. We show that CLIP significantly outperforms widely-used visual encoders trained with in-domain annotated data, such as BottomUp-TopDown. We achieve competitive or better results on diverse V&L tasks, while establishing new state-of-the-art results on Visual Question Answering, Visual Entailment, and V&L Navigation tasks. We release our code at https://github.com/clip-vil/CLIP-ViL.

We present a simple self-supervised method to enhance the performance of ViT features for dense downstream tasks. Our Lightweight Feature Transform (LiFT) is a straightforward and compact postprocessing network that can be applied to enhance the features of any pre-trained ViT backbone. LiFT is fast and easy to train with a self-supervised objective, and it boosts the density of ViT features for minimal extra inference cost. Furthermore, we demonstrate that LiFT can be applied with approaches that use additional task-specific downstream modules, as we integrate LiFT with ViTDet for COCO detection and segmentation. Despite the simplicity of LiFT, we find that it is not simply learning a more complex version of bilinear interpolation. Instead, our LiFT training protocol leads to several desirable emergent properties that benefit ViT features in dense downstream tasks. This includes greater scale invariance for features, and better object boundary maps. By simply training LiFT for a few epochs, we show improved performance on keypoint correspondence, detection, segmentation, and object discovery tasks. Overall, LiFT provides an easy way to unlock the benefits of denser feature arrays for a fraction of the computational cost. For more details, refer to our project page at https://www.cs.umd.edu/~sakshams/LiFT/.

The success of contrastive language-image pretraining (CLIP) relies on the supervision from the pairing between images and captions, which tends to be noisy in web-crawled data. We present Mixture of Data Experts (MoDE) and learn a system of CLIP data experts via clustering. Each data expert is trained on one data cluster, being less sensitive to false negative noises in other clusters. At inference time, we ensemble their outputs by applying weights determined through the correlation between task metadata and cluster conditions. To estimate the correlation precisely, the samples in one cluster should be semantically similar, but the number of data experts should still be reasonable for training and inference. As such, we consider the ontology in human language and propose to use fine-grained cluster centers to represent each data expert at a coarse-grained level. Experimental studies show that four CLIP data experts on ViT-B/16 outperform the ViT-L/14 by OpenAI CLIP and OpenCLIP on zero-shot image classification but with less (<35\%) training cost. Meanwhile, MoDE can train all data expert asynchronously and can flexibly include new data experts. The code is available at https://github.com/facebookresearch/MetaCLIP/tree/main/mode.

Contrastive Language-Image Pre-training (CLIP) is an approach that has advanced research and applications in computer vision, fueling modern recognition systems and generative models. We believe that the main ingredient to the success of CLIP is its data and not the model architecture or pre-training objective. However, CLIP only provides very limited information about its data and how it has been collected, leading to works that aim to reproduce CLIP's data by filtering with its model parameters. In this work, we intend to reveal CLIP's data curation approach and in our pursuit of making it open to the community introduce Metadata-Curated Language-Image Pre-training (MetaCLIP). MetaCLIP takes a raw data pool and metadata (derived from CLIP's concepts) and yields a balanced subset over the metadata distribution. Our experimental study rigorously isolates the model and training settings, concentrating solely on data. MetaCLIP applied to CommonCrawl with 400M image-text data pairs outperforms CLIP's data on multiple standard benchmarks. In zero-shot ImageNet classification, MetaCLIP achieves 70.8% accuracy, surpassing CLIP's 68.3% on ViT-B models. Scaling to 1B data, while maintaining the same training budget, attains 72.4%. Our observations hold across various model sizes, exemplified by ViT-H achieving 80.5%, without any bells-and-whistles. Curation code and training data distribution on metadata is made available at https://github.com/facebookresearch/MetaCLIP.

Contrastive Language-Image Pretraining (CLIP) stands out as a prominent method for image representation learning. Various neural architectures, spanning Transformer-based models like Vision Transformers (ViTs) to Convolutional Networks (ConvNets) like ResNets, are trained with CLIP and serve as universal backbones across diverse vision tasks. Despite utilizing the same data and training objectives, the effectiveness of representations learned by these architectures raises a critical question. Our investigation explores the differences in CLIP performance among these backbone architectures, revealing significant disparities in their classifications. Notably, normalizing these representations results in substantial performance variations. Our findings showcase a remarkable possible synergy between backbone predictions that could reach an improvement of over 20% through informed selection of the appropriate backbone. Moreover, we propose a simple, yet effective approach to combine predictions from multiple backbones, leading to a notable performance boost of up to 6.34\%. We will release the code for reproducing the results.

Vision transformers (ViTs) are quickly becoming the de-facto architecture for computer vision, yet we understand very little about why they work and what they learn. While existing studies visually analyze the mechanisms of convolutional neural networks, an analogous exploration of ViTs remains challenging. In this paper, we first address the obstacles to performing visualizations on ViTs. Assisted by these solutions, we observe that neurons in ViTs trained with language model supervision (e.g., CLIP) are activated by semantic concepts rather than visual features. We also explore the underlying differences between ViTs and CNNs, and we find that transformers detect image background features, just like their convolutional counterparts, but their predictions depend far less on high-frequency information. On the other hand, both architecture types behave similarly in the way features progress from abstract patterns in early layers to concrete objects in late layers. In addition, we show that ViTs maintain spatial information in all layers except the final layer. In contrast to previous works, we show that the last layer most likely discards the spatial information and behaves as a learned global pooling operation. Finally, we conduct large-scale visualizations on a wide range of ViT variants, including DeiT, CoaT, ConViT, PiT, Swin, and Twin, to validate the effectiveness of our method.

In the absence of parallax cues, a learning-based single image depth estimation (SIDE) model relies heavily on shading and contextual cues in the image. While this simplicity is attractive, it is necessary to train such models on large and varied datasets, which are difficult to capture. It has been shown that using embeddings from pre-trained foundational models, such as CLIP, improves zero shot transfer in several applications. Taking inspiration from this, in our paper we explore the use of global image priors generated from a pre-trained ViT model to provide more detailed contextual information. We argue that the embedding vector from a ViT model, pre-trained on a large dataset, captures greater relevant information for SIDE than the usual route of generating pseudo image captions, followed by CLIP based text embeddings. Based on this idea, we propose a new SIDE model using a diffusion backbone which is conditioned on ViT embeddings. Our proposed design establishes a new state-of-the-art (SOTA) for SIDE on NYUv2 dataset, achieving Abs Rel error of 0.059 (14% improvement) compared to 0.069 by the current SOTA (VPD). And on KITTI dataset, achieving Sq Rel error of 0.139 (2% improvement) compared to 0.142 by the current SOTA (GEDepth). For zero-shot transfer with a model trained on NYUv2, we report mean relative improvement of (20%, 23%, 81%, 25%) over NeWCRFs on (Sun-RGBD, iBims1, DIODE, HyperSim) datasets, compared to (16%, 18%, 45%, 9%) by ZoeDepth. The project page is available at https://ecodepth-iitd.github.io

The Vision Transformer (ViT) has made significant strides in the field of computer vision. However, as the depth of the model and the resolution of the input images increase, the computational cost associated with training and running ViT models has surged dramatically. This paper proposes a hybrid model based on CNN and Vision Transformer, named CI2P-ViT. The model incorporates a module called CI2P, which utilizes the CompressAI encoder to compress images and subsequently generates a sequence of patches through a series of convolutions. CI2P can replace the Patch Embedding component in the ViT model, enabling seamless integration into existing ViT models. Compared to ViT-B/16, CI2P-ViT has the number of patches input to the self-attention layer reduced to a quarter of the original. This design not only significantly reduces the computational cost of the ViT model but also effectively enhances the model's accuracy by introducing the inductive bias properties of CNN. The ViT model's precision is markedly enhanced. When trained from the ground up on the Animals-10 dataset, CI2P-ViT achieved an accuracy rate of 92.37%, representing a 3.3% improvement over the ViT-B/16 baseline. Additionally, the model's computational operations, measured in floating-point operations per second (FLOPs), were diminished by 63.35%, and it exhibited a 2-fold increase in training velocity on identical hardware configurations.

Attention-based vision models, such as Vision Transformer (ViT) and its variants, have shown promising performance in various computer vision tasks. However, these emerging architectures suffer from large model sizes and high computational costs, calling for efficient model compression solutions. To date, pruning ViTs has been well studied, while other compression strategies that have been widely applied in CNN compression, e.g., model factorization, is little explored in the context of ViT compression. This paper explores an efficient method for compressing vision transformers to enrich the toolset for obtaining compact attention-based vision models. Based on the new insight on the multi-head attention layer, we develop a highly efficient ViT compression solution, which outperforms the state-of-the-art pruning methods. For compressing DeiT-small and DeiT-base models on ImageNet, our proposed approach can achieve 0.45% and 0.76% higher top-1 accuracy even with fewer parameters. Our finding can also be applied to improve the customization efficiency of text-to-image diffusion models, with much faster training (up to 2.6times speedup) and lower extra storage cost (up to 1927.5times reduction) than the existing works.

This paper tackles the challenge of detecting partially manipulated facial deepfakes, which involve subtle alterations to specific facial features while retaining the overall context, posing a greater detection difficulty than fully synthetic faces. We leverage the Contrastive Language-Image Pre-training (CLIP) model, specifically its ViT-L/14 visual encoder, to develop a generalizable detection method that performs robustly across diverse datasets and unknown forgery techniques with minimal modifications to the original model. The proposed approach utilizes parameter-efficient fine-tuning (PEFT) techniques, such as LN-tuning, to adjust a small subset of the model's parameters, preserving CLIP's pre-trained knowledge and reducing overfitting. A tailored preprocessing pipeline optimizes the method for facial images, while regularization strategies, including L2 normalization and metric learning on a hyperspherical manifold, enhance generalization. Trained on the FaceForensics++ dataset and evaluated in a cross-dataset fashion on Celeb-DF-v2, DFDC, FFIW, and others, the proposed method achieves competitive detection accuracy comparable to or outperforming much more complex state-of-the-art techniques. This work highlights the efficacy of CLIP's visual encoder in facial deepfake detection and establishes a simple, powerful baseline for future research, advancing the field of generalizable deepfake detection. The code is available at: https://github.com/yermandy/deepfake-detection

This paper investigates the capability of plain Vision Transformers (ViTs) for semantic segmentation using the encoder-decoder framework and introduces SegViTv2. In this study, we introduce a novel Attention-to-Mask (\atm) module to design a lightweight decoder effective for plain ViT. The proposed ATM converts the global attention map into semantic masks for high-quality segmentation results. Our decoder outperforms the popular decoder UPerNet using various ViT backbones while consuming only about 5% of the computational cost. For the encoder, we address the concern of the relatively high computational cost in the ViT-based encoders and propose a Shrunk++ structure that incorporates edge-aware query-based down-sampling (EQD) and query-based upsampling (QU) modules. The Shrunk++ structure reduces the computational cost of the encoder by up to 50% while maintaining competitive performance. Furthermore, we propose to adapt SegViT for continual semantic segmentation, demonstrating nearly zero forgetting of previously learned knowledge. Experiments show that our proposed SegViTv2 surpasses recent segmentation methods on three popular benchmarks including ADE20k, COCO-Stuff-10k and PASCAL-Context datasets. The code is available through the following link: https://github.com/zbwxp/SegVit.

Shift equivariance is a fundamental principle that governs how we perceive the world - our recognition of an object remains invariant with respect to shifts. Transformers have gained immense popularity due to their effectiveness in both language and vision tasks. While the self-attention operator in vision transformers (ViT) is permutation-equivariant and thus shift-equivariant, patch embedding, positional encoding, and subsampled attention in ViT variants can disrupt this property, resulting in inconsistent predictions even under small shift perturbations. Although there is a growing trend in incorporating the inductive bias of convolutional neural networks (CNNs) into vision transformers, it does not fully address the issue. We propose an adaptive polyphase anchoring algorithm that can be seamlessly integrated into vision transformer models to ensure shift-equivariance in patch embedding and subsampled attention modules, such as window attention and global subsampled attention. Furthermore, we utilize depth-wise convolution to encode positional information. Our algorithms enable ViT, and its variants such as Twins to achieve 100% consistency with respect to input shift, demonstrate robustness to cropping, flipping, and affine transformations, and maintain consistent predictions even when the original models lose 20 percentage points on average when shifted by just a few pixels with Twins' accuracy dropping from 80.57% to 62.40%.

Large-scale multi-modal training with image-text pairs imparts strong generalization to CLIP model. Since training on a similar scale for videos is infeasible, recent approaches focus on the effective transfer of image-based CLIP to the video domain. In this pursuit, new parametric modules are added to learn temporal information and inter-frame relationships which require meticulous design efforts. Furthermore, when the resulting models are learned on videos, they tend to overfit on the given task distribution and lack in generalization aspect. This begs the following question: How to effectively transfer image-level CLIP representations to videos? In this work, we show that a simple Video Fine-tuned CLIP (ViFi-CLIP) baseline is generally sufficient to bridge the domain gap from images to videos. Our qualitative analysis illustrates that the frame-level processing from CLIP image-encoder followed by feature pooling and similarity matching with corresponding text embeddings helps in implicitly modeling the temporal cues within ViFi-CLIP. Such fine-tuning helps the model to focus on scene dynamics, moving objects and inter-object relationships. For low-data regimes where full fine-tuning is not viable, we propose a `bridge and prompt' approach that first uses fine-tuning to bridge the domain gap and then learns prompts on language and vision side to adapt CLIP representations. We extensively evaluate this simple yet strong baseline on zero-shot, base-to-novel generalization, few-shot and fully supervised settings across five video benchmarks. Our code is available at https://github.com/muzairkhattak/ViFi-CLIP.

Vision Transformers (ViTs) have achieved remarkable success in image recognition, yet standard ViT architectures are hampered by substantial parameter redundancy and high computational cost, limiting their practical deployment. While recent efforts on efficient ViTs primarily focus on static model compression or token-level sparsification, they remain constrained by fixed computational depth for all tokens. In this work, we present MoR-ViT, a novel vision transformer framework that, for the first time, incorporates a token-level dynamic recursion mechanism inspired by the Mixture-of-Recursions (MoR) paradigm. This approach enables each token to adaptively determine its processing depth, yielding a flexible and input-dependent allocation of computational resources. Extensive experiments on ImageNet-1K and transfer benchmarks demonstrate that MoR-ViT not only achieves state-of-the-art accuracy with up to 70% parameter reduction and 2.5x inference acceleration, but also outperforms leading efficient ViT baselines such as DynamicViT and TinyViT under comparable conditions. These results establish dynamic recursion as an effective strategy for efficient vision transformers and open new avenues for scalable and deployable deep learning models in real-world scenarios.

We explore the application of Vision Transformer (ViT) for handwritten text recognition. The limited availability of labeled data in this domain poses challenges for achieving high performance solely relying on ViT. Previous transformer-based models required external data or extensive pre-training on large datasets to excel. To address this limitation, we introduce a data-efficient ViT method that uses only the encoder of the standard transformer. We find that incorporating a Convolutional Neural Network (CNN) for feature extraction instead of the original patch embedding and employ Sharpness-Aware Minimization (SAM) optimizer to ensure that the model can converge towards flatter minima and yield notable enhancements. Furthermore, our introduction of the span mask technique, which masks interconnected features in the feature map, acts as an effective regularizer. Empirically, our approach competes favorably with traditional CNN-based models on small datasets like IAM and READ2016. Additionally, it establishes a new benchmark on the LAM dataset, currently the largest dataset with 19,830 training text lines. The code is publicly available at: https://github.com/YutingLi0606/HTR-VT.

Vision Transformers (ViT) have marked a paradigm shift in computer vision, outperforming state-of-the-art models across diverse tasks. However, their practical deployment is hampered by high computational and memory demands. This study addresses the challenge by evaluating four primary model compression techniques: quantization, low-rank approximation, knowledge distillation, and pruning. We methodically analyze and compare the efficacy of these techniques and their combinations in optimizing ViTs for resource-constrained environments. Our comprehensive experimental evaluation demonstrates that these methods facilitate a balanced compromise between model accuracy and computational efficiency, paving the way for wider application in edge computing devices.

Visual encoding followed by token condensing has become the standard architectural paradigm in multi-modal large language models (MLLMs). Many recent MLLMs increasingly favor global native- resolution visual encoding over slice-based methods. To investigate this trend, we systematically compare their behavior on vision-language understanding and attention patterns, revealing that global encoding enhances overall capability but at the expense of greater computational overhead. To address this issue, we present LLaVA-UHD v3, an MLLM centered upon our proposed Progressive Visual Compression (PVC) method, which can be seamlessly integrated into standard Vision Transformer (ViT) to enable efficient native-resolution encoding. The PVC approach consists of two key modules: (i) refined patch embedding, which supports flexible patch-size scaling for fine-grained visual model- ing, (ii) windowed token compression, hierarchically deployed across ViT layers to progressively aggregate local token representations. Jointly modulated by these two modules, a widely pretrained ViT can be reconfigured into an efficient architecture while largely preserving generality. Evaluated across extensive benchmarks, the transformed ViT, termed ViT-UHD, demonstrates competitive performance with MoonViT while reducing TTFT (time-to-first-token) by 2.4x, when developed within an identical MLLM architecture. Building upon ViT-UHD, LLaVA-UHD v3 also achieves competitive performance to Qwen2-VL, while further reducing TTFT by 1.9x. We will release all code and checkpoints to support future research on efficient MLLMs.

Scale is the primary factor for building a powerful foundation model that could well generalize to a variety of downstream tasks. However, it is still challenging to train video foundation models with billions of parameters. This paper shows that video masked autoencoder (VideoMAE) is a scalable and general self-supervised pre-trainer for building video foundation models. We scale the VideoMAE in both model and data with a core design. Specifically, we present a dual masking strategy for efficient pre-training, with an encoder operating on a subset of video tokens and a decoder processing another subset of video tokens. Although VideoMAE is very efficient due to high masking ratio in encoder, masking decoder can still further reduce the overall computational cost. This enables the efficient pre-training of billion-level models in video. We also use a progressive training paradigm that involves an initial pre-training on a diverse multi-sourced unlabeled dataset, followed by a post-pre-training on a mixed labeled dataset. Finally, we successfully train a video ViT model with a billion parameters, which achieves a new state-of-the-art performance on the datasets of Kinetics (90.0% on K400 and 89.9% on K600) and Something-Something (68.7% on V1 and 77.0% on V2). In addition, we extensively verify the pre-trained video ViT models on a variety of downstream tasks, demonstrating its effectiveness as a general video representation learner. The code and model is available at https://github.com/OpenGVLab/VideoMAEv2.

Despite the remarkable success of Vision Transformers (ViTs) in various visual tasks, they are often hindered by substantial computational cost. In this work, we introduce Vote\&Mix (VoMix), a plug-and-play and parameter-free token reduction method, which can be readily applied to off-the-shelf ViT models without any training. VoMix tackles the computational redundancy of ViTs by identifying tokens with high homogeneity through a layer-wise token similarity voting mechanism. Subsequently, the selected tokens are mixed into the retained set, thereby preserving visual information. Experiments demonstrate VoMix significantly improves the speed-accuracy tradeoff of ViTs on both images and videos. Without any training, VoMix achieves a 2times increase in throughput of existing ViT-H on ImageNet-1K and a 2.4times increase in throughput of existing ViT-L on Kinetics-400 video dataset, with a mere 0.3\% drop in top-1 accuracy.

This work presents a simple vision transformer design as a strong baseline for object localization and instance segmentation tasks. Transformers recently demonstrate competitive performance in image classification tasks. To adopt ViT to object detection and dense prediction tasks, many works inherit the multistage design from convolutional networks and highly customized ViT architectures. Behind this design, the goal is to pursue a better trade-off between computational cost and effective aggregation of multiscale global contexts. However, existing works adopt the multistage architectural design as a black-box solution without a clear understanding of its true benefits. In this paper, we comprehensively study three architecture design choices on ViT -- spatial reduction, doubled channels, and multiscale features -- and demonstrate that a vanilla ViT architecture can fulfill this goal without handcrafting multiscale features, maintaining the original ViT design philosophy. We further complete a scaling rule to optimize our model's trade-off on accuracy and computation cost / model size. By leveraging a constant feature resolution and hidden size throughout the encoder blocks, we propose a simple and compact ViT architecture called Universal Vision Transformer (UViT) that achieves strong performance on COCO object detection and instance segmentation tasks.

Vision-language pre-training like CLIP has shown promising performance on various downstream tasks such as zero-shot image classification and image-text retrieval. Most of the existing CLIP-alike works usually adopt relatively large image encoders like ResNet50 and ViT, while the lightweight counterparts are rarely discussed. In this paper, we propose a multi-level interaction paradigm for training lightweight CLIP models. Firstly, to mitigate the problem that some image-text pairs are not strictly one-to-one correspondence, we improve the conventional global instance-level alignment objective by softening the label of negative samples progressively. Secondly, a relaxed bipartite matching based token-level alignment objective is introduced for finer-grained alignment between image patches and textual words. Moreover, based on the observation that the accuracy of CLIP model does not increase correspondingly as the parameters of text encoder increase, an extra objective of masked language modeling (MLM) is leveraged for maximizing the potential of the shortened text encoder. In practice, an auxiliary fusion module injecting unmasked image embedding into masked text embedding at different network stages is proposed for enhancing the MLM. Extensive experiments show that without introducing additional computational cost during inference, the proposed method achieves a higher performance on multiple downstream tasks.

This paper explores feature prediction as a stand-alone objective for unsupervised learning from video and introduces V-JEPA, a collection of vision models trained solely using a feature prediction objective, without the use of pretrained image encoders, text, negative examples, reconstruction, or other sources of supervision. The models are trained on 2 million videos collected from public datasets and are evaluated on downstream image and video tasks. Our results show that learning by predicting video features leads to versatile visual representations that perform well on both motion and appearance-based tasks, without adaption of the model's parameters; e.g., using a frozen backbone. Our largest model, a ViT-H/16 trained only on videos, obtains 81.9% on Kinetics-400, 72.2% on Something-Something-v2, and 77.9% on ImageNet1K.

Medical image segmentation remains a formidable challenge due to the label scarcity. Pre-training Vision Transformer (ViT) through masked image modeling (MIM) on large-scale unlabeled medical datasets presents a promising solution, providing both computational efficiency and model generalization for various downstream tasks. However, current ViT-based MIM pre-training frameworks predominantly emphasize local aggregation representations in output layers and fail to exploit the rich representations across different ViT layers that better capture fine-grained semantic information needed for more precise medical downstream tasks. To fill the above gap, we hereby present Hierarchical Encoder-driven MAE (Hi-End-MAE), a simple yet effective ViT-based pre-training solution, which centers on two key innovations: (1) Encoder-driven reconstruction, which encourages the encoder to learn more informative features to guide the reconstruction of masked patches; and (2) Hierarchical dense decoding, which implements a hierarchical decoding structure to capture rich representations across different layers. We pre-train Hi-End-MAE on a large-scale dataset of 10K CT scans and evaluated its performance across seven public medical image segmentation benchmarks. Extensive experiments demonstrate that Hi-End-MAE achieves superior transfer learning capabilities across various downstream tasks, revealing the potential of ViT in medical imaging applications. The code is available at: https://github.com/FengheTan9/Hi-End-MAE

Contrastive Language-Image Pretraining (CLIP) models are able to capture the semantic relationship of images and texts and have enabled a wide range of applications, from image retrieval to classification. These models are trained with datasets extracted from web crawls, which are of large quantity but limited quality. This paper explores whether limited amounts higher quality data in a specific domain improve the general performance of CLIP models. To this purpose, we extract text-image data from scientific papers hosted in the arXiv and PubMed Central repositories. Experiments on small-scale CLIP models (ViT B/32) show that model performance increases on average, but only moderately. This result indicates that using the data sources considered in the paper to train large-scale CLIP models is a worthwile research direction.

Vision-language models (VLMs) like CLIP have shown impressive zero-shot and few-shot learning capabilities across diverse applications. However, adapting these models to new fine-grained domains remains difficult due to reliance on prompt engineering and the high cost of full model fine-tuning. Existing adaptation approaches rely on augmented components, such as prompt tokens and adapter modules, which could limit adaptation quality, destabilize the model, and compromise the rich knowledge learned during pretraining. In this work, we present CLIP-SVD, a novel multi-modal and parameter-efficient adaptation technique that leverages Singular Value Decomposition (SVD) to modify the internal parameter space of CLIP without injecting additional modules. Specifically, we fine-tune only the singular values of the CLIP parameter matrices to rescale the basis vectors for domain adaptation while retaining the pretrained model. This design enables enhanced adaptation performance using only 0.04\% of the model's total parameters and better preservation of its generalization ability. CLIP-SVD achieves state-of-the-art classification results on 11 natural and 10 biomedical datasets, outperforming previous methods in both accuracy and generalization under few-shot settings. Additionally, we leverage a natural language-based approach to analyze the effectiveness and dynamics of the CLIP adaptation to allow interpretability of CLIP-SVD. The code is publicly available at https://github.com/HealthX-Lab/CLIP-SVD.

Vision Transformer (ViT) demonstrates that Transformer for natural language processing can be applied to computer vision tasks and result in comparable performance to convolutional neural networks (CNN), which have been studied and adopted in computer vision for years. This naturally raises the question of how the performance of ViT can be advanced with design techniques of CNN. To this end, we propose to incorporate two techniques and present ViT-ResNAS, an efficient multi-stage ViT architecture designed with neural architecture search (NAS). First, we propose residual spatial reduction to decrease sequence lengths for deeper layers and utilize a multi-stage architecture. When reducing lengths, we add skip connections to improve performance and stabilize training deeper networks. Second, we propose weight-sharing NAS with multi-architectural sampling. We enlarge a network and utilize its sub-networks to define a search space. A super-network covering all sub-networks is then trained for fast evaluation of their performance. To efficiently train the super-network, we propose to sample and train multiple sub-networks with one forward-backward pass. After that, evolutionary search is performed to discover high-performance network architectures. Experiments on ImageNet demonstrate that ViT-ResNAS achieves better accuracy-MACs and accuracy-throughput trade-offs than the original DeiT and other strong baselines of ViT. Code is available at https://github.com/yilunliao/vit-search.

OpenAI's CLIP, released in early 2021, have long been the go-to choice of vision encoder for building multimodal foundation models. Although recent alternatives such as SigLIP have begun to challenge this status quo, to our knowledge none are fully open: their training data remains proprietary and/or their training recipes are not released. This paper fills this gap with OpenVision, a fully-open, cost-effective family of vision encoders that match or surpass the performance of OpenAI's CLIP when integrated into multimodal frameworks like LLaVA. OpenVision builds on existing works -- e.g., CLIPS for training framework and Recap-DataComp-1B for training data -- while revealing multiple key insights in enhancing encoder quality and showcasing practical benefits in advancing multimodal models. By releasing vision encoders spanning from 5.9M to 632.1M parameters, OpenVision offers practitioners a flexible trade-off between capacity and efficiency in building multimodal models: larger models deliver enhanced multimodal performance, while smaller versions enable lightweight, edge-ready multimodal deployments.

Recent advances in Vision Transformer (ViT) have demonstrated its impressive performance in image classification, which makes it a promising alternative to Convolutional Neural Network (CNN). Unlike CNNs, ViT represents an input image as a sequence of image patches. The patch-based input image representation makes the following question interesting: How does ViT perform when individual input image patches are perturbed with natural corruptions or adversarial perturbations, compared to CNNs? In this work, we study the robustness of ViT to patch-wise perturbations. Surprisingly, we find that ViTs are more robust to naturally corrupted patches than CNNs, whereas they are more vulnerable to adversarial patches. Furthermore, we discover that the attention mechanism greatly affects the robustness of vision transformers. Specifically, the attention module can help improve the robustness of ViT by effectively ignoring natural corrupted patches. However, when ViTs are attacked by an adversary, the attention mechanism can be easily fooled to focus more on the adversarially perturbed patches and cause a mistake. Based on our analysis, we propose a simple temperature-scaling based method to improve the robustness of ViT against adversarial patches. Extensive qualitative and quantitative experiments are performed to support our findings, understanding, and improvement of ViT robustness to patch-wise perturbations across a set of transformer-based architectures.

Generating and editing images from open domain text prompts is a challenging task that heretofore has required expensive and specially trained models. We demonstrate a novel methodology for both tasks which is capable of producing images of high visual quality from text prompts of significant semantic complexity without any training by using a multimodal encoder to guide image generations. We demonstrate on a variety of tasks how using CLIP [37] to guide VQGAN [11] produces higher visual quality outputs than prior, less flexible approaches like DALL-E [38], GLIDE [33] and Open-Edit [24], despite not being trained for the tasks presented. Our code is available in a public repository.

Multi-modal foundation models such as CLIP have showcased impressive zero-shot capabilities. However, their applicability in resource-constrained environments is limited due to their large number of parameters and high inference time. While existing approaches have scaled down the entire CLIP architecture, we focus on training smaller variants of the image encoder, which suffices for efficient zero-shot classification. The use of synthetic data has shown promise in distilling representations from larger teachers, resulting in strong few-shot and linear probe performance. However, we find that this approach surprisingly fails in true zero-shot settings when using contrastive losses. We identify the exploitation of spurious features as being responsible for poor generalization between synthetic and real data. However, by using the image feature-based L2 distillation loss, we mitigate these problems and train students that achieve zero-shot performance which on four domain-specific datasets is on-par with a ViT-B/32 teacher model trained on DataCompXL, while featuring up to 92% fewer parameters.

Vision Transformers (ViTs) have emerged as the dominant architecture for visual processing tasks, demonstrating excellent scalability with increased training data and model size. However, recent work has identified the emergence of artifact tokens in ViTs that are incongruous with the local semantics. These anomalous tokens degrade ViT performance in tasks that require fine-grained localization or structural coherence. An effective mitigation of this issue is to the addition of register tokens to ViTs, which implicitly "absorb" the artifact term during training. Given the availability of various large-scale pre-trained ViTs, in this paper we aim at equipping them with such register tokens without the need of re-training them from scratch, which is infeasible considering their size. Specifically, we propose Post Hoc Registers (PH-Reg), an efficient self-distillation method that integrates registers into an existing ViT without requiring additional labeled data and full retraining. PH-Reg initializes both teacher and student networks from the same pre-trained ViT. The teacher remains frozen and unmodified, while the student is augmented with randomly initialized register tokens. By applying test-time augmentation to the teacher's inputs, we generate denoised dense embeddings free of artifacts, which are then used to optimize only a small subset of unlocked student weights. We show that our approach can effectively reduce the number of artifact tokens, improving the segmentation and depth prediction of the student ViT under zero-shot and linear probing.

Image token removal is an efficient augmentation strategy for reducing the cost of computing image features. However, this efficient augmentation strategy has been found to adversely affect the accuracy of CLIP-based training. We hypothesize that removing a large portion of image tokens may improperly discard the semantic content associated with a given text description, thus constituting an incorrect pairing target in CLIP training. To address this issue, we propose an attentive token removal approach for CLIP training, which retains tokens with a high semantic correlation to the text description. The correlation scores are computed in an online fashion using the EMA version of the visual encoder. Our experiments show that the proposed attentive masking approach performs better than the previous method of random token removal for CLIP training. The approach also makes it efficient to apply multiple augmentation views to the image, as well as introducing instance contrastive learning tasks between these views into the CLIP framework. Compared to other CLIP improvements that combine different pre-training targets such as SLIP and MaskCLIP, our method is not only more effective, but also much more efficient. Specifically, using ViT-B and YFCC-15M dataset, our approach achieves 43.9% top-1 accuracy on ImageNet-1K zero-shot classification, as well as 62.7/42.1 and 38.0/23.2 I2T/T2I retrieval accuracy on Flickr30K and MS COCO, which are +1.1%, +5.5/+0.9, and +4.4/+1.3 higher than the SLIP method, while being 2.30times faster. An efficient version of our approach running 1.16times faster than the plain CLIP model achieves significant gains of +5.3%, +11.3/+8.0, and +9.5/+4.9 on these benchmarks.

Several recent studies have demonstrated that attention-based networks, such as Vision Transformer (ViT), can outperform Convolutional Neural Networks (CNNs) on several computer vision tasks without using convolutional layers. This naturally leads to the following questions: Can a self-attention layer of ViT express any convolution operation? In this work, we prove that a single ViT layer with image patches as the input can perform any convolution operation constructively, where the multi-head attention mechanism and the relative positional encoding play essential roles. We further provide a lower bound on the number of heads for Vision Transformers to express CNNs. Corresponding with our analysis, experimental results show that the construction in our proof can help inject convolutional bias into Transformers and significantly improve the performance of ViT in low data regimes.

We present an approach to efficiently and effectively adapt a masked image modeling (MIM) pre-trained vanilla Vision Transformer (ViT) for object detection, which is based on our two novel observations: (i) A MIM pre-trained vanilla ViT encoder can work surprisingly well in the challenging object-level recognition scenario even with randomly sampled partial observations, e.g., only 25% sim 50% of the input embeddings. (ii) In order to construct multi-scale representations for object detection from single-scale ViT, a randomly initialized compact convolutional stem supplants the pre-trained large kernel patchify stem, and its intermediate features can naturally serve as the higher resolution inputs of a feature pyramid network without further upsampling or other manipulations. While the pre-trained ViT is only regarded as the 3^{rd}-stage of our detector's backbone instead of the whole feature extractor. This results in a ConvNet-ViT hybrid feature extractor. The proposed detector, named MIMDet, enables a MIM pre-trained vanilla ViT to outperform hierarchical Swin Transformer by 2.5 box AP and 2.6 mask AP on COCO, and achieves better results compared with the previous best adapted vanilla ViT detector using a more modest fine-tuning recipe while converging 2.8times faster. Code and pre-trained models are available at https://github.com/hustvl/MIMDet.

The recently proposed SparseFormer architecture provides an alternative approach to visual understanding by utilizing a significantly lower number of visual tokens via adjusting RoIs, greatly reducing computational costs while still achieving promising performance. However, training SparseFormers from scratch is still expensive, and scaling up the number of parameters can be challenging. In this paper, we propose to bootstrap SparseFormers from ViT-based vision foundation models in a simple and efficient way. Since the majority of SparseFormer blocks are the standard transformer ones, we can inherit weights from large-scale pre-trained vision transformers and freeze them as much as possible. Therefore, we only need to train the SparseFormer-specific lightweight focusing transformer to adjust token RoIs and fine-tune a few early pre-trained blocks to align the final token representation. In such a way, we can bootstrap SparseFormer architectures from various large-scale pre-trained models (e.g., IN-21K pre-trained AugRegs or CLIPs) using a rather smaller amount of training samples (e.g., IN-1K) and without labels or captions within just a few hours. As a result, the bootstrapped unimodal SparseFormer (from AugReg-ViT-L/16-384) can reach 84.9% accuracy on IN-1K with only 49 tokens, and the multimodal SparseFormer from CLIPs also demonstrates notable zero-shot performance with highly reduced computational cost without seeing any caption during the bootstrapping procedure. In addition, CLIP-bootstrapped SparseFormers, which align the output space with language without seeing a word, can serve as efficient vision encoders in multimodal large language models. Code will be publicly available at https://github.com/showlab/sparseformer

Vision Language Models (VLMs) like CLIP have attracted substantial attention in pathology, serving as backbones for applications such as zero-shot image classification and Whole Slide Image (WSI) analysis. Additionally, they can function as vision encoders when combined with large language models (LLMs) to support broader capabilities. Current efforts to train pathology VLMs rely on pathology image-text pairs from platforms like PubMed, YouTube, and Twitter, which provide limited, unscalable data with generally suboptimal image quality. In this work, we leverage large-scale WSI datasets like TCGA to extract numerous high-quality image patches. We then train a large multimodal model to generate captions for these images, creating PathGen-1.6M, a dataset containing 1.6 million high-quality image-caption pairs. Our approach involves multiple agent models collaborating to extract representative WSI patches, generating and refining captions to obtain high-quality image-text pairs. Extensive experiments show that integrating these generated pairs with existing datasets to train a pathology-specific CLIP model, PathGen-CLIP, significantly enhances its ability to analyze pathological images, with substantial improvements across nine pathology-related zero-shot image classification tasks and three whole-slide image tasks. Furthermore, we construct 200K instruction-tuning data based on PathGen-1.6M and integrate PathGen-CLIP with the Vicuna LLM to create more powerful multimodal models through instruction tuning. Overall, we provide a scalable pathway for high-quality data generation in pathology, paving the way for next-generation general pathology models.

Scaling up contrastive language-image pretraining (CLIP) is critical for empowering both vision and multimodal models. We present EVA-CLIP-18B, the largest and most powerful open-source CLIP model to date, with 18-billion parameters. With only 6-billion training samples seen, EVA-CLIP-18B achieves an exceptional 80.7% zero-shot top-1 accuracy averaged across 27 widely recognized image classification benchmarks, outperforming its forerunner EVA-CLIP (5-billion parameters) and other open-source CLIP models by a large margin. Remarkably, we observe a consistent performance improvement with the model size scaling of EVA-CLIP, despite maintaining a constant training dataset of 2-billion image-text pairs from LAION-2B and COYO-700M. This dataset is openly available and much smaller than the in-house datasets (e.g., DFN-5B, WebLI-10B) employed in other state-of-the-art CLIP models. EVA-CLIP-18B demonstrates the potential of EVA-style weak-to-strong visual model scaling. With our model weights made publicly available, we hope to facilitate future research in vision and multimodal foundation models.

While vision transformers (ViTs) have shown great potential in computer vision tasks, their intense computation and memory requirements pose challenges for practical applications. Existing post-training quantization methods leverage value redistribution or specialized quantizers to address the non-normal distribution in ViTs. However, without considering the asymmetry in activations and relying on hand-crafted settings, these methods often struggle to maintain performance under low-bit quantization. To overcome these challenges, we introduce SmoothQuant with bias term (SQ-b) to alleviate the asymmetry issue and reduce the clamping loss. We also introduce optimal scaling factor ratio search (OPT-m) to determine quantization parameters by a data-dependent mechanism automatically. To further enhance the compressibility, we incorporate the above-mentioned techniques and propose a mixed-precision post-training quantization framework for vision transformers (MPTQ-ViT). We develop greedy mixed-precision quantization (Greedy MP) to allocate layer-wise bit-width considering both model performance and compressibility. Our experiments on ViT, DeiT, and Swin demonstrate significant accuracy improvements compared with SOTA on the ImageNet dataset. Specifically, our proposed methods achieve accuracy improvements ranging from 0.90% to 23.35% on 4-bit ViTs with single-precision and from 3.82% to 78.14% on 5-bit fully quantized ViTs with mixed-precision.

Large pretrained plain vision Transformers (ViTs) have been the workhorse for many downstream tasks. However, existing works utilizing off-the-shelf ViTs are inefficient in terms of training and deployment, because adopting ViTs with individual sizes requires separate trainings and is restricted by fixed performance-efficiency trade-offs. In this paper, we are inspired by stitchable neural networks (SN-Net), which is a new framework that cheaply produces a single model that covers rich subnetworks by stitching pretrained model families, supporting diverse performance-efficiency trade-offs at runtime. Building upon this foundation, we introduce SN-Netv2, a systematically improved model stitching framework to facilitate downstream task adaptation. Specifically, we first propose a two-way stitching scheme to enlarge the stitching space. We then design a resource-constrained sampling strategy that takes into account the underlying FLOPs distributions in the space for better sampling. Finally, we observe that learning stitching layers as a low-rank update plays an essential role on downstream tasks to stabilize training and ensure a good Pareto frontier. With extensive experiments on ImageNet-1K, ADE20K, COCO-Stuff-10K and NYUv2, SN-Netv2 demonstrates superior performance over SN-Netv1 on downstream dense predictions and shows strong ability as a flexible vision backbone, achieving great advantages in both training efficiency and deployment flexibility. Code is available at https://github.com/ziplab/SN-Netv2.

Contrastive Language-Image Pre-training (CLIP) models have shown significant potential, particularly in zero-shot classification across diverse distribution shifts. Building on existing evaluations of overall classification robustness, this work aims to provide a more comprehensive assessment of CLIP by introducing several new perspectives. First, we investigate their robustness to variations in specific visual factors. Second, we assess two critical safety objectives--confidence uncertainty and out-of-distribution detection--beyond mere classification accuracy. Third, we evaluate the finesse with which CLIP models bridge the image and text modalities. Fourth, we extend our examination to 3D awareness in CLIP models, moving beyond traditional 2D image understanding. Finally, we explore the interaction between vision and language encoders within modern large multimodal models (LMMs) that utilize CLIP as the visual backbone, focusing on how this interaction impacts classification robustness. In each aspect, we consider the impact of six factors on CLIP models: model architecture, training distribution, training set size, fine-tuning, contrastive loss, and test-time prompts. Our study uncovers several previously unknown insights into CLIP. For instance, the architecture of the visual encoder in CLIP plays a significant role in their robustness against 3D corruption. CLIP models tend to exhibit a bias towards shape when making predictions. Moreover, this bias tends to diminish after fine-tuning on ImageNet. Vision-language models like LLaVA, leveraging the CLIP vision encoder, could exhibit benefits in classification performance for challenging categories over CLIP alone. Our findings are poised to offer valuable guidance for enhancing the robustness and reliability of CLIP models.

Albeit the scalable performance of vision transformers (ViTs), the dense computational costs (training & inference) undermine their position in industrial applications. Post-training quantization (PTQ), tuning ViTs with a tiny dataset and running in a low-bit format, well addresses the cost issue but unluckily bears more performance drops in lower-bit cases. In this paper, we introduce I&S-ViT, a novel method that regulates the PTQ of ViTs in an inclusive and stable fashion. I&S-ViT first identifies two issues in the PTQ of ViTs: (1) Quantization inefficiency in the prevalent log2 quantizer for post-Softmax activations; (2) Rugged and magnified loss landscape in coarse-grained quantization granularity for post-LayerNorm activations. Then, I&S-ViT addresses these issues by introducing: (1) A novel shift-uniform-log2 quantizer (SULQ) that incorporates a shift mechanism followed by uniform quantization to achieve both an inclusive domain representation and accurate distribution approximation; (2) A three-stage smooth optimization strategy (SOS) that amalgamates the strengths of channel-wise and layer-wise quantization to enable stable learning. Comprehensive evaluations across diverse vision tasks validate I&S-ViT' superiority over existing PTQ of ViTs methods, particularly in low-bit scenarios. For instance, I&S-ViT elevates the performance of 3-bit ViT-B by an impressive 50.68%.

Vision Transformer (ViT) has emerged as a powerful architecture in the realm of modern computer vision. However, its application in certain imaging fields, such as microscopy and satellite imaging, presents unique challenges. In these domains, images often contain multiple channels, each carrying semantically distinct and independent information. Furthermore, the model must demonstrate robustness to sparsity in input channels, as they may not be densely available during training or testing. In this paper, we propose a modification to the ViT architecture that enhances reasoning across the input channels and introduce Hierarchical Channel Sampling (HCS) as an additional regularization technique to ensure robustness when only partial channels are presented during test time. Our proposed model, ChannelViT, constructs patch tokens independently from each input channel and utilizes a learnable channel embedding that is added to the patch tokens, similar to positional embeddings. We evaluate the performance of ChannelViT on ImageNet, JUMP-CP (microscopy cell imaging), and So2Sat (satellite imaging). Our results show that ChannelViT outperforms ViT on classification tasks and generalizes well, even when a subset of input channels is used during testing. Across our experiments, HCS proves to be a powerful regularizer, independent of the architecture employed, suggesting itself as a straightforward technique for robust ViT training. Lastly, we find that ChannelViT generalizes effectively even when there is limited access to all channels during training, highlighting its potential for multi-channel imaging under real-world conditions with sparse sensors. Our code is available at https://github.com/insitro/ChannelViT.

his paper tackles a significant challenge faced by Vision Transformers (ViTs): their constrained scalability across different image resolutions. Typically, ViTs experience a performance decline when processing resolutions different from those seen during training. Our work introduces two key innovations to address this issue. Firstly, we propose a novel module for dynamic resolution adjustment, designed with a single Transformer block, specifically to achieve highly efficient incremental token integration. Secondly, we introduce fuzzy positional encoding in the Vision Transformer to provide consistent positional awareness across multiple resolutions, thereby preventing overfitting to any single training resolution. Our resulting model, ViTAR (Vision Transformer with Any Resolution), demonstrates impressive adaptability, achieving 83.3\% top-1 accuracy at a 1120x1120 resolution and 80.4\% accuracy at a 4032x4032 resolution, all while reducing computational costs. ViTAR also shows strong performance in downstream tasks such as instance and semantic segmentation and can easily combined with self-supervised learning techniques like Masked AutoEncoder. Our work provides a cost-effective solution for enhancing the resolution scalability of ViTs, paving the way for more versatile and efficient high-resolution image processing.

Vision transformers (ViTs) have emerged as a significant area of focus, particularly for their capacity to be jointly trained with large language models and to serve as robust vision foundation models. Yet, the development of trustworthy explanation methods for ViTs has lagged, particularly in the context of post-hoc interpretations of ViT predictions. Existing sub-image selection approaches, such as feature-attribution and conceptual models, fall short in this regard. This paper proposes five desiderata for explaining ViTs -- faithfulness, stability, sparsity, multi-level structure, and parsimony -- and demonstrates the inadequacy of current methods in meeting these criteria comprehensively. We introduce a variational Bayesian explanation framework, dubbed ProbAbilistic Concept Explainers (PACE), which models the distributions of patch embeddings to provide trustworthy post-hoc conceptual explanations. Our qualitative analysis reveals the distributions of patch-level concepts, elucidating the effectiveness of ViTs by modeling the joint distribution of patch embeddings and ViT's predictions. Moreover, these patch-level explanations bridge the gap between image-level and dataset-level explanations, thus completing the multi-level structure of PACE. Through extensive experiments on both synthetic and real-world datasets, we demonstrate that PACE surpasses state-of-the-art methods in terms of the defined desiderata.

The success of Vision Transformer (ViT) has been widely reported on a wide range of image recognition tasks. ViT can learn global dependencies superior to CNN, yet CNN's inherent locality can substitute for expensive training resources. Recently, the outstanding performance of RetNet in the field of language modeling has garnered attention, surpassing that of the Transformer with explicit local modeling, shifting researchers' focus towards Transformers in the CV field. This paper investigates the effectiveness of RetNet from a CNN perspective and presents a variant of RetNet tailored to the visual domain. Similar to RetNet we improves ViT's local modeling by applying a weight mask on the original self-attention matrix. A straightforward way to locally adapt the self-attention matrix can be realized by an element-wise learnable weight mask (ELM), for which our preliminary results show promising results. However, the element-wise simple learnable weight mask not only induces a non-trivial additional parameter overhead but also increases the optimization complexity. To this end, this work proposes a novel Gaussian mixture mask (GMM) in which one mask only has two learnable parameters and it can be conveniently used in any ViT variants whose attention mechanism allows the use of masks. Experimental results on multiple small datasets demonstrate that the effectiveness of our proposed Gaussian mask for boosting ViTs for free (almost zero additional parameter or computation cost). Our code can be publicly available at https://github.com/CatworldLee/Gaussian-Mixture-Mask-Attention.

The synergy between generative and discriminative models receives growing attention. While discriminative Contrastive Language-Image Pre-Training (CLIP) excels in high-level semantics, it struggles with perceiving fine-grained visual details. Generally, to enhance representations, generative models take CLIP's visual features as conditions for reconstruction. However, the underlying principle remains underexplored. In this work, we empirically found that visually perfect generations are not always optimal for representation enhancement. The essence lies in effectively extracting fine-grained knowledge from generative models while mitigating irrelevant information. To explore critical factors, we delve into three aspects: (1) Conditioning mechanisms: We found that even a small number of local tokens can drastically reduce the difficulty of reconstruction, leading to collapsed training. We thus conclude that utilizing only global visual tokens as conditions is the most effective strategy. (2) Denoising configurations: We observed that end-to-end training introduces extraneous information. To address this, we propose a two-stage training strategy to prioritize learning useful visual knowledge. Additionally, we demonstrate that lightweight denoisers can yield remarkable improvements. (3) Generation paradigms: We explore both continuous and discrete denoisers with desirable outcomes, validating the versatility of our method. Through our in-depth explorations, we have finally arrived at an effective method, namely GenHancer, which consistently outperforms prior arts on the MMVP-VLM benchmark, e.g., 6.0% on OpenAICLIP. The enhanced CLIP can be further plugged into multimodal large language models for better vision-centric performance. All the models and codes are made publicly available.

Contrastive Language-Image Pre-training (CLIP) exhibits strong zero-shot classification ability on various image-level tasks, leading to the research to adapt CLIP for pixel-level open-vocabulary semantic segmentation without additional training. The key is to improve spatial representation of image-level CLIP, such as replacing self-attention map at last layer with self-self attention map or vision foundation model based attention map. In this paper, we present a novel hierarchical framework, named CLIPer, that hierarchically improves spatial representation of CLIP. The proposed CLIPer includes an early-layer fusion module and a fine-grained compensation module. We observe that, the embeddings and attention maps at early layers can preserve spatial structural information. Inspired by this, we design the early-layer fusion module to generate segmentation map with better spatial coherence. Afterwards, we employ a fine-grained compensation module to compensate the local details using the self-attention maps of diffusion model. We conduct the experiments on seven segmentation datasets. Our proposed CLIPer achieves the state-of-the-art performance on these datasets. For instance, using ViT-L, CLIPer has the mIoU of 69.8% and 43.3% on VOC and COCO Object, outperforming ProxyCLIP by 9.2% and 4.1% respectively.

Structural re-parameterization is a general training scheme for Convolutional Neural Networks (CNNs), which achieves performance improvement without increasing inference cost. As Vision Transformers (ViTs) are gradually surpassing CNNs in various visual tasks, one may question: if a training scheme specifically for ViTs exists that can also achieve performance improvement without increasing inference cost? Recently, Mixture-of-Experts (MoE) has attracted increasing attention, as it can efficiently scale up the capacity of Transformers at a fixed cost through sparsely activated experts. Considering that MoE can also be viewed as a multi-branch structure, can we utilize MoE to implement a ViT training scheme similar to structural re-parameterization? In this paper, we affirmatively answer these questions, with a new general training strategy for ViTs. Specifically, we decouple the training and inference phases of ViTs. During training, we replace some Feed-Forward Networks (FFNs) of the ViT with specially designed, more efficient MoEs that assign tokens to experts by random uniform partition, and perform Experts Weights Averaging (EWA) on these MoEs at the end of each iteration. After training, we convert each MoE into an FFN by averaging the experts, transforming the model back into original ViT for inference. We further provide a theoretical analysis to show why and how it works. Comprehensive experiments across various 2D and 3D visual tasks, ViT architectures, and datasets validate the effectiveness and generalizability of the proposed training scheme. Besides, our training scheme can also be applied to improve performance when fine-tuning ViTs. Lastly, but equally important, the proposed EWA technique can significantly improve the effectiveness of naive MoE in various 2D visual small datasets and 3D visual tasks.

We propose a new transformer-based image and video tokenizer with Binary Spherical Quantization (BSQ). BSQ projects the high-dimensional visual embedding to a lower-dimensional hypersphere and then applies binary quantization. BSQ is (1) parameter-efficient without an explicit codebook, (2) scalable to arbitrary token dimensions, and (3) compact: compressing visual data by up to 100times with minimal distortion. Our tokenizer uses a transformer encoder and decoder with simple block-wise causal masking to support variable-length videos as input. The resulting BSQ-ViT achieves state-of-the-art visual reconstruction quality on image and video reconstruction benchmarks with 2.4times throughput compared to the best prior methods. Furthermore, by learning an autoregressive prior for adaptive arithmetic coding, BSQ-ViT achieves comparable results on video compression with state-of-the-art video compression standards. BSQ-ViT also enables masked language models to achieve competitive image synthesis quality to GAN- and diffusion-based methods.

The parameter-efficient adaptation of the image-text pretraining model CLIP for video-text retrieval is a prominent area of research. While CLIP is focused on image-level vision-language matching, video-text retrieval demands comprehensive understanding at the video level. Three key discrepancies emerge in the transfer from image-level to video-level: vision, language, and alignment. However, existing methods mainly focus on vision while neglecting language and alignment. In this paper, we propose Discrepancy Reduction in Vision, Language, and Alignment (DiscoVLA), which simultaneously mitigates all three discrepancies. Specifically, we introduce Image-Video Features Fusion to integrate image-level and video-level features, effectively tackling both vision and language discrepancies. Additionally, we generate pseudo image captions to learn fine-grained image-level alignment. To mitigate alignment discrepancies, we propose Image-to-Video Alignment Distillation, which leverages image-level alignment knowledge to enhance video-level alignment. Extensive experiments demonstrate the superiority of our DiscoVLA. In particular, on MSRVTT with CLIP (ViT-B/16), DiscoVLA outperforms previous methods by 1.5% in R@1, reaching a final score of 50.5% R@1. The code is available at https://github.com/LunarShen/DsicoVLA.

CLIP (Contrastive Language-Image Pre-training) models developed by OpenAI have achieved outstanding results on various image recognition and retrieval tasks, displaying strong zero-shot performance. This means that they are able to perform effectively on tasks for which they have not been explicitly trained. Inspired by the success of OpenAI CLIP, a new publicly available dataset called LAION-5B was collected which resulted in the development of open ViT-H/14, ViT-G/14 models that outperform the OpenAI L/14 model. The LAION-5B dataset also released an approximate nearest neighbor index, with a web interface for search & subset creation. In this paper, we evaluate the performance of various CLIP models as zero-shot face recognizers. Our findings show that CLIP models perform well on face recognition tasks, but increasing the size of the CLIP model does not necessarily lead to improved accuracy. Additionally, we investigate the robustness of CLIP models against data poisoning attacks by testing their performance on poisoned data. Through this analysis, we aim to understand the potential consequences and misuse of search engines built using CLIP models, which could potentially function as unintentional face recognition engines.

This paper provides a simplification on OpenVision's architecture and loss design for enhancing its training efficiency. Following the prior vision-language pretraining works CapPa and AIMv2, as well as modern multimodal designs like LLaVA, our changes are straightforward: we remove the text encoder (and therefore the contrastive loss), retaining only the captioning loss as a purely generative training signal. We name this new version OpenVision 2. The initial results are promising: despite this simplification, OpenVision 2 competitively matches the original model's performance on a broad set of multimodal benchmarks while substantially cutting both training time and memory consumption. For example, with ViT-L/14, it reduces training time by about 1.5x (from 83h to 57h), and memory usage by about 1.8x (from 24.5GB to 13.8GB, equivalently allowing the maximum batch size to grow from 2k to 8k). This superior training efficiency also allows us to scale far beyond the largest vision encoder used in OpenVision, reaching more than 1 billion parameters. We hold a strong belief that this lightweight, generative-only paradigm is compelling for future vision encoder development in multimodal foundation models.

Large-scale but noisy image-text pair data have paved the way for the success of Contrastive Language-Image Pretraining (CLIP). As the foundation vision encoder, CLIP in turn serves as the cornerstone for most large vision-language models (LVLMs). This interdependence naturally raises an interesting question: Can we reciprocally leverage LVLMs to enhance the quality of image-text pair data, thereby opening the possibility of a self-reinforcing cycle for continuous improvement? In this work, we take a significant step toward this vision by introducing an LVLM-driven data refinement pipeline. Our framework leverages LVLMs to process images and their raw alt-text, generating four complementary textual formulas: long positive descriptions, long negative descriptions, short positive tags, and short negative tags. Applying this pipeline to the curated DFN-Large dataset yields VLM-150M, a refined dataset enriched with multi-grained annotations. Based on this dataset, we further propose a training paradigm that extends conventional contrastive learning by incorporating negative descriptions and short tags as additional supervised signals. The resulting model, namely HQ-CLIP, demonstrates remarkable improvements across diverse benchmarks. Within a comparable training data scale, our approach achieves state-of-the-art performance in zero-shot classification, cross-modal retrieval, and fine-grained visual understanding tasks. In retrieval benchmarks, HQ-CLIP even surpasses standard CLIP models trained on the DFN-2B dataset, which contains 10times more training data than ours. All code, data, and models are available at https://zxwei.site/hqclip.

We present VPNeXt, a new and simple model for the Plain Vision Transformer (ViT). Unlike the many related studies that share the same homogeneous paradigms, VPNeXt offers a fresh perspective on dense representation based on ViT. In more detail, the proposed VPNeXt addressed two concerns about the existing paradigm: (1) Is it necessary to use a complex Transformer Mask Decoder architecture to obtain good representations? (2) Does the Plain ViT really need to depend on the mock pyramid feature for upsampling? For (1), we investigated the potential underlying reasons that contributed to the effectiveness of the Transformer Decoder and introduced the Visual Context Replay (VCR) to achieve similar effects efficiently. For (2), we introduced the ViTUp module. This module fully utilizes the previously overlooked ViT real pyramid feature to achieve better upsampling results compared to the earlier mock pyramid feature. This represents the first instance of such functionality in the field of semantic segmentation for Plain ViT. We performed ablation studies on related modules to verify their effectiveness gradually. We conducted relevant comparative experiments and visualizations to show that VPNeXt achieved state-of-the-art performance with a simple and effective design. Moreover, the proposed VPNeXt significantly exceeded the long-established mIoU wall/barrier of the VOC2012 dataset, setting a new state-of-the-art by a large margin, which also stands as the largest improvement since 2015.

We introduce new methods for 1) accelerating and 2) stabilizing training for large language-vision models. 1) For acceleration, we introduce SwitchBack, a linear layer for int8 quantized training which provides a speed-up of 13-25% while matching the performance of bfloat16 training within 0.1 percentage points for the 1B parameter CLIP ViT-Huge -- the largest int8 training to date. Our main focus is int8 as GPU support for float8 is rare, though we also analyze float8 training through simulation. While SwitchBack proves effective for float8, we show that standard techniques are also successful if the network is trained and initialized so that large feature magnitudes are discouraged, which we accomplish via layer-scale initialized with zeros. 2) For stability, we analyze loss spikes and find they consistently occur 1-8 iterations after the squared gradients become under-estimated by their AdamW second moment estimator. As a result, we recommend an AdamW-Adafactor hybrid which avoids loss spikes when training a CLIP ViT-Huge model and outperforms gradient clipping at the scales we test.

Pretrained vision-language models, such as CLIP, show promising zero-shot performance across a wide variety of datasets. For closed-set classification tasks, however, there is an inherent limitation: CLIP image encoders are typically designed to extract generic image-level features that summarize superfluous or confounding information for the target tasks. This results in degradation of classification performance, especially when objects of interest cover small areas of input images. In this work, we propose CLIP with Guided Cropping (GC-CLIP), where we use an off-the-shelf zero-shot object detection model in a preprocessing step to increase focus of zero-shot classifier to the object of interest and minimize influence of extraneous image regions. We empirically show that our approach improves zero-shot classification results across architectures and datasets, favorably for small objects.

We present a single neural network architecture composed of task-agnostic components (ViTs, convolutions, and LSTMs) that achieves state-of-art results on both the ImageNav ("go to location in <this picture>") and ObjectNav ("find a chair") tasks without any task-specific modules like object detection, segmentation, mapping, or planning modules. Such general-purpose methods offer advantages of simplicity in design, positive scaling with available compute, and versatile applicability to multiple tasks. Our work builds upon the recent success of self-supervised learning (SSL) for pre-training vision transformers (ViT). However, while the training recipes for convolutional networks are mature and robust, the recipes for ViTs are contingent and brittle, and in the case of ViTs for visual navigation, yet to be fully discovered. Specifically, we find that vanilla ViTs do not outperform ResNets on visual navigation. We propose the use of a compression layer operating over ViT patch representations to preserve spatial information along with policy training improvements. These improvements allow us to demonstrate positive scaling laws for the first time in visual navigation tasks. Consequently, our model advances state-of-the-art performance on ImageNav from 54.2% to 82.0% success and performs competitively against concurrent state-of-art on ObjectNav with success rate of 64.0% vs. 65.0%. Overall, this work does not present a fundamentally new approach, but rather recommendations for training a general-purpose architecture that achieves state-of-art performance today and could serve as a strong baseline for future methods.

Vision transformers are known to be more computationally and data-intensive than CNN models. These transformer models such as ViT, require all the input image tokens to learn the relationship among them. However, many of these tokens are not informative and may contain irrelevant information such as unrelated background or unimportant scenery. These tokens are overlooked by the multi-head self-attention (MHSA), resulting in many redundant and unnecessary computations in MHSA and the feed-forward network (FFN). In this work, we propose a method to optimize the amount of unnecessary interactions between unimportant tokens by separating and sending them through a different low-cost computational path. Our method does not add any parameters to the ViT model and aims to find the best trade-off between training throughput and achieving a 0% loss in the Top-1 accuracy of the final model. Our experimental results on training ViT-small from scratch show that SkipViT is capable of effectively dropping 55% of the tokens while gaining more than 13% training throughput and maintaining classification accuracy at the level of the baseline model on Huawei Ascend910A.

A Vision Transformer (ViT) is a simple neural architecture amenable to serve several computer vision tasks. It has limited built-in architectural priors, in contrast to more recent architectures that incorporate priors either about the input data or of specific tasks. Recent works show that ViTs benefit from self-supervised pre-training, in particular BerT-like pre-training like BeiT. In this paper, we revisit the supervised training of ViTs. Our procedure builds upon and simplifies a recipe introduced for training ResNet-50. It includes a new simple data-augmentation procedure with only 3 augmentations, closer to the practice in self-supervised learning. Our evaluations on Image classification (ImageNet-1k with and without pre-training on ImageNet-21k), transfer learning and semantic segmentation show that our procedure outperforms by a large margin previous fully supervised training recipes for ViT. It also reveals that the performance of our ViT trained with supervision is comparable to that of more recent architectures. Our results could serve as better baselines for recent self-supervised approaches demonstrated on ViT.

Generating fine-grained video descriptions is a fundamental challenge in video understanding. In this work, we introduce Tarsier, a family of large-scale video-language models designed to generate high-quality video descriptions. Tarsier employs CLIP-ViT to encode frames separately and then uses an LLM to model temporal relationships. Despite its simple architecture, we demonstrate that with a meticulously designed two-stage training procedure, the Tarsier models exhibit substantially stronger video description capabilities than any existing open-source model, showing a +51.4% advantage in human side-by-side evaluation over the strongest model. Additionally, they are comparable to state-of-the-art proprietary models, with a +12.3% advantage against GPT-4V and a -6.7% disadvantage against Gemini 1.5 Pro. Besides video description, Tarsier proves to be a versatile generalist model, achieving new state-of-the-art results across nine public benchmarks, including multi-choice VQA, open-ended VQA, and zero-shot video captioning. Our second contribution is the introduction of a new benchmark for evaluating video description models, consisting of a new challenging dataset featuring videos from diverse sources and varying complexity, along with an automatic method specifically designed to assess the quality of fine-grained video descriptions. We make our models and evaluation benchmark publicly available at https://github.com/bytedance/tarsier.

Pre-trained vision-language models~(VLMs) are the de-facto foundation models for various downstream tasks. However, scene text recognition methods still prefer backbones pre-trained on a single modality, namely, the visual modality, despite the potential of VLMs to serve as powerful scene text readers. For example, CLIP can robustly identify regular (horizontal) and irregular (rotated, curved, blurred, or occluded) text in images. With such merits, we transform CLIP into a scene text reader and introduce CLIP4STR, a simple yet effective STR method built upon image and text encoders of CLIP. It has two encoder-decoder branches: a visual branch and a cross-modal branch. The visual branch provides an initial prediction based on the visual feature, and the cross-modal branch refines this prediction by addressing the discrepancy between the visual feature and text semantics. To fully leverage the capabilities of both branches, we design a dual predict-and-refine decoding scheme for inference. We scale CLIP4STR in terms of the model size, pre-training data, and training data, achieving state-of-the-art performance on 11 STR benchmarks. Additionally, a comprehensive empirical study is provided to enhance the understanding of the adaptation of CLIP to STR. We believe our method establishes a simple yet strong baseline for future STR research with VLMs.

There still remains an extreme performance gap between Vision Transformers (ViTs) and Convolutional Neural Networks (CNNs) when training from scratch on small datasets, which is concluded to the lack of inductive bias. In this paper, we further consider this problem and point out two weaknesses of ViTs in inductive biases, that is, the spatial relevance and diverse channel representation. First, on spatial aspect, objects are locally compact and relevant, thus fine-grained feature needs to be extracted from a token and its neighbors. While the lack of data hinders ViTs to attend the spatial relevance. Second, on channel aspect, representation exhibits diversity on different channels. But the scarce data can not enable ViTs to learn strong enough representation for accurate recognition. To this end, we propose Dynamic Hybrid Vision Transformer (DHVT) as the solution to enhance the two inductive biases. On spatial aspect, we adopt a hybrid structure, in which convolution is integrated into patch embedding and multi-layer perceptron module, forcing the model to capture the token features as well as their neighboring features. On channel aspect, we introduce a dynamic feature aggregation module in MLP and a brand new "head token" design in multi-head self-attention module to help re-calibrate channel representation and make different channel group representation interacts with each other. The fusion of weak channel representation forms a strong enough representation for classification. With this design, we successfully eliminate the performance gap between CNNs and ViTs, and our DHVT achieves a series of state-of-the-art performance with a lightweight model, 85.68% on CIFAR-100 with 22.8M parameters, 82.3% on ImageNet-1K with 24.0M parameters. Code is available at https://github.com/ArieSeirack/DHVT.

With the growing interest in pretrained vision-language models like CLIP, recent research has focused on adapting these models to downstream tasks. Despite achieving promising results, most existing methods require labeled data for all classes, which may not hold in real-world applications due to the long tail and Zipf's law. For example, some classes may lack labeled data entirely, such as emerging concepts. To address this problem, we propose a plug-and-play generative approach called SyntHesIzed Prompts~(SHIP) to improve existing fine-tuning methods. Specifically, we follow variational autoencoders to introduce a generator that reconstructs the visual features by inputting the synthesized prompts and the corresponding class names to the textual encoder of CLIP. In this manner, we easily obtain the synthesized features for the remaining label-only classes. Thereafter, we fine-tune CLIP with off-the-shelf methods by combining labeled and synthesized features. Extensive experiments on base-to-new generalization, cross-dataset transfer learning, and generalized zero-shot learning demonstrate the superiority of our approach. The code is available at https://github.com/mrflogs/SHIP.

Understanding the memorization and privacy leakage risks in Contrastive Language--Image Pretraining (CLIP) is critical for ensuring the security of multimodal models. Recent studies have demonstrated the feasibility of extracting sensitive training examples from diffusion models, with conditional diffusion models exhibiting a stronger tendency to memorize and leak information. In this work, we investigate data memorization and extraction risks in CLIP through the lens of CLIP inversion, a process that aims to reconstruct training images from text prompts. To this end, we introduce LeakyCLIP, a novel attack framework designed to achieve high-quality, semantically accurate image reconstruction from CLIP embeddings. We identify three key challenges in CLIP inversion: 1) non-robust features, 2) limited visual semantics in text embeddings, and 3) low reconstruction fidelity. To address these challenges, LeakyCLIP employs 1) adversarial fine-tuning to enhance optimization smoothness, 2) linear transformation-based embedding alignment, and 3) Stable Diffusion-based refinement to improve fidelity. Empirical results demonstrate the superiority of LeakyCLIP, achieving over 358% improvement in Structural Similarity Index Measure (SSIM) for ViT-B-16 compared to baseline methods on LAION-2B subset. Furthermore, we uncover a pervasive leakage risk, showing that training data membership can even be successfully inferred from the metrics of low-fidelity reconstructions. Our work introduces a practical method for CLIP inversion while offering novel insights into the nature and scope of privacy risks in multimodal models.

We consider the problem of adapting a contrastively pretrained vision-language model like CLIP (Radford et al., 2021) for few-shot classification. The existing literature addresses this problem by learning a linear classifier of the frozen visual features, optimizing word embeddings, or learning external feature adapters. This paper introduces an alternative way for CLIP adaptation without adding 'external' parameters to optimize. We find that simply fine-tuning the last projection matrix of the vision encoder leads to strong performance compared to the existing baselines. Furthermore, we show that regularizing training with the distance between the fine-tuned and pretrained matrices adds reliability for adapting CLIP through this layer. Perhaps surprisingly, this approach, coined ProLIP, yields performances on par or better than state of the art on 11 few-shot classification benchmarks, few-shot domain generalization, cross-dataset transfer and test-time adaptation. Code will be made available at https://github.com/astra-vision/ProLIP .

Vision Transformer (ViT) has become one of the most popular neural architectures due to its great scalability, computational efficiency, and compelling performance in many vision tasks. However, ViT has shown inferior performance to Convolutional Neural Network (CNN) on medical tasks due to its data-hungry nature and the lack of annotated medical data. In this paper, we pre-train ViTs on 266,340 chest X-rays using Masked Autoencoders (MAE) which reconstruct missing pixels from a small part of each image. For comparison, CNNs are also pre-trained on the same 266,340 X-rays using advanced self-supervised methods (e.g., MoCo v2). The results show that our pre-trained ViT performs comparably (sometimes better) to the state-of-the-art CNN (DenseNet-121) for multi-label thorax disease classification. This performance is attributed to the strong recipes extracted from our empirical studies for pre-training and fine-tuning ViT. The pre-training recipe signifies that medical reconstruction requires a much smaller proportion of an image (10% vs. 25%) and a more moderate random resized crop range (0.5~1.0 vs. 0.2~1.0) compared with natural imaging. Furthermore, we remark that in-domain transfer learning is preferred whenever possible. The fine-tuning recipe discloses that layer-wise LR decay, RandAug magnitude, and DropPath rate are significant factors to consider. We hope that this study can direct future research on the application of Transformers to a larger variety of medical imaging tasks.

Vision Transformers (ViTs) mark a revolutionary advance in neural networks with their token mixer's powerful global context capability. However, the pairwise token affinity and complex matrix operations limit its deployment on resource-constrained scenarios and real-time applications, such as mobile devices, although considerable efforts have been made in previous works. In this paper, we introduce CAS-ViT: Convolutional Additive Self-attention Vision Transformers, to achieve a balance between efficiency and performance in mobile applications. Firstly, we argue that the capability of token mixers to obtain global contextual information hinges on multiple information interactions, such as spatial and channel domains. Subsequently, we construct a novel additive similarity function following this paradigm and present an efficient implementation named Convolutional Additive Token Mixer (CATM). This simplification leads to a significant reduction in computational overhead. We evaluate CAS-ViT across a variety of vision tasks, including image classification, object detection, instance segmentation, and semantic segmentation. Our experiments, conducted on GPUs, ONNX, and iPhones, demonstrate that CAS-ViT achieves a competitive performance when compared to other state-of-the-art backbones, establishing it as a viable option for efficient mobile vision applications. Our code and model are available at: https://github.com/Tianfang-Zhang/CAS-ViT

The recent work CLIPA presents an inverse scaling law for CLIP training -- whereby the larger the image/text encoders used, the shorter the sequence length of image/text tokens that can be applied in training. This finding enables us to train high-performance CLIP models with significantly reduced computations. Building upon this work, we hereby present CLIPA-v2 with two key contributions. Technically, we find this inverse scaling law is also applicable in the finetuning stage, enabling further reduction in computational needs. Empirically, we explore CLIPA at scale, extending the experiments up to the H/14 model with ~13B image-text pairs seen during training. Our results are exciting -- by only allocating a budget of \10,000, our CLIP model achieves an impressive zero-shot ImageNet accuracy of 81.1%, surpassing the prior best CLIP model (from OpenCLIP, 80.1%) by 1.0% and meanwhile reducing the computational cost by ~39X. Moreover, with an additional investment of 4,000, we can further elevate the zero-shot ImageNet accuracy to 81.8%. Our code and models are available at https://github.com/UCSC-VLAA/CLIPA.

This paper introduces DeeCLIP, a novel framework for detecting AI-generated images using CLIP-ViT and fusion learning. Despite significant advancements in generative models capable of creating highly photorealistic images, existing detection methods often struggle to generalize across different models and are highly sensitive to minor perturbations. To address these challenges, DeeCLIP incorporates DeeFuser, a fusion module that combines high-level and low-level features, improving robustness against degradations such as compression and blurring. Additionally, we apply triplet loss to refine the embedding space, enhancing the model's ability to distinguish between real and synthetic content. To further enable lightweight adaptation while preserving pre-trained knowledge, we adopt parameter-efficient fine-tuning using low-rank adaptation (LoRA) within the CLIP-ViT backbone. This approach supports effective zero-shot learning without sacrificing generalization. Trained exclusively on 4-class ProGAN data, DeeCLIP achieves an average accuracy of 89.00% on 19 test subsets composed of generative adversarial network (GAN) and diffusion models. Despite having fewer trainable parameters, DeeCLIP outperforms existing methods, demonstrating superior robustness against various generative models and real-world distortions. The code is publicly available at https://github.com/Mamadou-Keita/DeeCLIP for research purposes.

In this paper, we address the challenge of making ViT models more robust to unseen affine transformations. Such robustness becomes useful in various recognition tasks such as face recognition when image alignment failures occur. We propose a novel method called KP-RPE, which leverages key points (e.g.~facial landmarks) to make ViT more resilient to scale, translation, and pose variations. We begin with the observation that Relative Position Encoding (RPE) is a good way to bring affine transform generalization to ViTs. RPE, however, can only inject the model with prior knowledge that nearby pixels are more important than far pixels. Keypoint RPE (KP-RPE) is an extension of this principle, where the significance of pixels is not solely dictated by their proximity but also by their relative positions to specific keypoints within the image. By anchoring the significance of pixels around keypoints, the model can more effectively retain spatial relationships, even when those relationships are disrupted by affine transformations. We show the merit of KP-RPE in face and gait recognition. The experimental results demonstrate the effectiveness in improving face recognition performance from low-quality images, particularly where alignment is prone to failure. Code and pre-trained models are available.

Despite the success of Vision-Language Models (VLMs) like CLIP in aligning vision and language, their proficiency in detailed, fine-grained visual comprehension remains a key challenge. We present CLIP-IN, a novel framework that bolsters CLIP's fine-grained perception through two core innovations. Firstly, we leverage instruction-editing datasets, originally designed for image manipulation, as a unique source of hard negative image-text pairs. Coupled with a symmetric hard negative contrastive loss, this enables the model to effectively distinguish subtle visual-semantic differences. Secondly, CLIP-IN incorporates long descriptive captions, utilizing rotary positional encodings to capture rich semantic context often missed by standard CLIP. Our experiments demonstrate that CLIP-IN achieves substantial gains on the MMVP benchmark and various fine-grained visual recognition tasks, without compromising robust zero-shot performance on broader classification and retrieval tasks. Critically, integrating CLIP-IN's visual representations into Multimodal Large Language Models significantly reduces visual hallucinations and enhances reasoning abilities. This work underscores the considerable potential of synergizing targeted, instruction-based contrastive learning with comprehensive descriptive information to elevate the fine-grained understanding of VLMs.

Although convolutional networks have been the dominant architecture for vision tasks for many years, recent experiments have shown that Transformer-based models, most notably the Vision Transformer (ViT), may exceed their performance in some settings. However, due to the quadratic runtime of the self-attention layers in Transformers, ViTs require the use of patch embeddings, which group together small regions of the image into single input features, in order to be applied to larger image sizes. This raises a question: Is the performance of ViTs due to the inherently-more-powerful Transformer architecture, or is it at least partly due to using patches as the input representation? In this paper, we present some evidence for the latter: specifically, we propose the ConvMixer, an extremely simple model that is similar in spirit to the ViT and the even-more-basic MLP-Mixer in that it operates directly on patches as input, separates the mixing of spatial and channel dimensions, and maintains equal size and resolution throughout the network. In contrast, however, the ConvMixer uses only standard convolutions to achieve the mixing steps. Despite its simplicity, we show that the ConvMixer outperforms the ViT, MLP-Mixer, and some of their variants for similar parameter counts and data set sizes, in addition to outperforming classical vision models such as the ResNet. Our code is available at https://github.com/locuslab/convmixer.

Vision-language models (VLMs) are typically composed of a vision encoder, e.g. CLIP, and a language model (LM) that interprets the encoded features to solve downstream tasks. Despite remarkable progress, VLMs are subject to several shortcomings due to the limited capabilities of vision encoders, e.g. "blindness" to certain image features, visual hallucination, etc. To address these issues, we study broadening the visual encoding capabilities of VLMs. We first comprehensively benchmark several vision encoders with different inductive biases for solving VLM tasks. We observe that there is no single encoding configuration that consistently achieves top performance across different tasks, and encoders with different biases can perform surprisingly similarly. Motivated by this, we introduce a method, named BRAVE, that consolidates features from multiple frozen encoders into a more versatile representation that can be directly fed as the input to a frozen LM. BRAVE achieves state-of-the-art performance on a broad range of captioning and VQA benchmarks and significantly reduces the aforementioned issues of VLMs, while requiring a smaller number of trainable parameters than existing methods and having a more compressed representation. Our results highlight the potential of incorporating different visual biases for a more broad and contextualized visual understanding of VLMs.

Transformers have recently emerged as a powerful tool for learning visual representations. In this paper, we identify and characterize artifacts in feature maps of both supervised and self-supervised ViT networks. The artifacts correspond to high-norm tokens appearing during inference primarily in low-informative background areas of images, that are repurposed for internal computations. We propose a simple yet effective solution based on providing additional tokens to the input sequence of the Vision Transformer to fill that role. We show that this solution fixes that problem entirely for both supervised and self-supervised models, sets a new state of the art for self-supervised visual models on dense visual prediction tasks, enables object discovery methods with larger models, and most importantly leads to smoother feature maps and attention maps for downstream visual processing.

Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have led to significant progress in 2D body pose estimation. However, achieving a good balance between accuracy, efficiency, and robustness remains a challenge. For instance, CNNs are computationally efficient but struggle with long-range dependencies, while ViTs excel in capturing such dependencies but suffer from quadratic computational complexity. This paper proposes two ViT-based models for accurate, efficient, and robust 2D pose estimation. The first one, EViTPose, operates in a computationally efficient manner without sacrificing accuracy by utilizing learnable joint tokens to select and process a subset of the most important body patches, enabling us to control the trade-off between accuracy and efficiency by changing the number of patches to be processed. The second one, UniTransPose, while not allowing for the same level of direct control over the trade-off, efficiently handles multiple scales by combining (1) an efficient multi-scale transformer encoder that uses both local and global attention with (2) an efficient sub-pixel CNN decoder for better speed and accuracy. Moreover, by incorporating all joints from different benchmarks into a unified skeletal representation, we train robust methods that learn from multiple datasets simultaneously and perform well across a range of scenarios -- including pose variations, lighting conditions, and occlusions. Experiments on six benchmarks demonstrate that the proposed methods significantly outperform state-of-the-art methods while improving computational efficiency. EViTPose exhibits a significant decrease in computational complexity (30% to 44% less in GFLOPs) with a minimal drop of accuracy (0% to 3.5% less), and UniTransPose achieves accuracy improvements ranging from 0.9% to 43.8% across these benchmarks.

Vision Transformers (ViTs) have gained significant popularity in recent years and have proliferated into many applications. However, their behavior under different learning paradigms is not well explored. We compare ViTs trained through different methods of supervision, and show that they learn a diverse range of behaviors in terms of their attention, representations, and downstream performance. We also discover ViT behaviors that are consistent across supervision, including the emergence of Offset Local Attention Heads. These are self-attention heads that attend to a token adjacent to the current token with a fixed directional offset, a phenomenon that to the best of our knowledge has not been highlighted in any prior work. Our analysis shows that ViTs are highly flexible and learn to process local and global information in different orders depending on their training method. We find that contrastive self-supervised methods learn features that are competitive with explicitly supervised features, and they can even be superior for part-level tasks. We also find that the representations of reconstruction-based models show non-trivial similarity to contrastive self-supervised models. Project website (https://www.cs.umd.edu/~sakshams/vit_analysis) and code (https://www.github.com/mwalmer-umd/vit_analysis) are publicly available.

This paper does not describe a novel method. Instead, it studies a straightforward, incremental, yet must-know baseline given the recent progress in computer vision: self-supervised learning for Vision Transformers (ViT). While the training recipes for standard convolutional networks have been highly mature and robust, the recipes for ViT are yet to be built, especially in the self-supervised scenarios where training becomes more challenging. In this work, we go back to basics and investigate the effects of several fundamental components for training self-supervised ViT. We observe that instability is a major issue that degrades accuracy, and it can be hidden by apparently good results. We reveal that these results are indeed partial failure, and they can be improved when training is made more stable. We benchmark ViT results in MoCo v3 and several other self-supervised frameworks, with ablations in various aspects. We discuss the currently positive evidence as well as challenges and open questions. We hope that this work will provide useful data points and experience for future research.

Vision transformers have emerged as a promising alternative to convolutional neural networks for various image analysis tasks, offering comparable or superior performance. However, one significant drawback of ViTs is their resource-intensive nature, leading to increased memory footprint, computation complexity, and power consumption. To democratize this high-performance technology and make it more environmentally friendly, it is essential to compress ViT models, reducing their resource requirements while maintaining high performance. In this paper, we introduce a new block-structured pruning to address the resource-intensive issue for ViTs, offering a balanced trade-off between accuracy and hardware acceleration. Unlike unstructured pruning or channel-wise structured pruning, block pruning leverages the block-wise structure of linear layers, resulting in more efficient matrix multiplications. To optimize this pruning scheme, our paper proposes a novel hardware-aware learning objective that simultaneously maximizes speedup and minimizes power consumption during inference, tailored to the block sparsity structure. This objective eliminates the need for empirical look-up tables and focuses solely on reducing parametrized layer connections. Moreover, our paper provides a lightweight algorithm to achieve post-training pruning for ViTs, utilizing second-order Taylor approximation and empirical optimization to solve the proposed hardware-aware objective. Extensive experiments on ImageNet are conducted across various ViT architectures, including DeiT-B and DeiT-S, demonstrating competitive performance with other pruning methods and achieving a remarkable balance between accuracy preservation and power savings. Especially, we achieve up to 3.93x and 1.79x speedups on dedicated hardware and GPUs respectively for DeiT-B, and also observe an inference power reduction by 1.4x on real-world GPUs.

CLIP is a discriminative model trained to align images and text in a shared embedding space. Due to its multimodal structure, it serves as the backbone of many generative pipelines, where a decoder is trained to map from the shared space back to images. In this work, we show that image synthesis is nevertheless possible using CLIP alone -- without any decoder, training, or fine-tuning. Our approach optimizes a frequency-aware implicit neural representation that encourages coarse-to-fine generation by stratifying frequencies across network layers. To stabilize this inverse mapping, we introduce adversarially robust initialization, a lightweight Orthogonal Procrustes projection to align local text and image embeddings, and a blending loss that anchors outputs to natural image statistics. Without altering CLIP's weights, this framework unlocks capabilities such as text-to-image generation, style transfer, and image reconstruction. These findings suggest that discriminative models may hold untapped generative potential, hidden in plain sight.

Vision Transformers (ViTs) have shown impressive performance and have become a unified backbone for multiple vision tasks. But both attention and multi-layer perceptions (MLPs) in ViTs are not efficient enough due to dense multiplications, resulting in costly training and inference. To this end, we propose to reparameterize the pre-trained ViT with a mixture of multiplication primitives, e.g., bitwise shifts and additions, towards a new type of multiplication-reduced model, dubbed ShiftAddViT, which aims for end-to-end inference speedups on GPUs without the need of training from scratch. Specifically, all MatMuls among queries, keys, and values are reparameterized by additive kernels, after mapping queries and keys to binary codes in Hamming space. The remaining MLPs or linear layers are then reparameterized by shift kernels. We utilize TVM to implement and optimize those customized kernels for practical hardware deployment on GPUs. We find that such a reparameterization on (quadratic or linear) attention maintains model accuracy, while inevitably leading to accuracy drops when being applied to MLPs. To marry the best of both worlds, we further propose a new mixture of experts (MoE) framework to reparameterize MLPs by taking multiplication or its primitives as experts, e.g., multiplication and shift, and designing a new latency-aware load-balancing loss. Such a loss helps to train a generic router for assigning a dynamic amount of input tokens to different experts according to their latency. In principle, the faster experts run, the larger amount of input tokens are assigned. Extensive experiments consistently validate the effectiveness of our proposed ShiftAddViT, achieving up to 5.18\times$ latency reductions on GPUs and 42.9%$ energy savings, while maintaining comparable accuracy as original or efficient ViTs.

In recent computer vision research, the advent of the Vision Transformer (ViT) has rapidly revolutionized various architectural design efforts: ViT achieved state-of-the-art image classification performance using self-attention found in natural language processing, and MLP-Mixer achieved competitive performance using simple multi-layer perceptrons. In contrast, several studies have also suggested that carefully redesigned convolutional neural networks (CNNs) can achieve advanced performance comparable to ViT without resorting to these new ideas. Against this background, there is growing interest in what inductive bias is suitable for computer vision. Here we propose Sequencer, a novel and competitive architecture alternative to ViT that provides a new perspective on these issues. Unlike ViTs, Sequencer models long-range dependencies using LSTMs rather than self-attention layers. We also propose a two-dimensional version of Sequencer module, where an LSTM is decomposed into vertical and horizontal LSTMs to enhance performance. Despite its simplicity, several experiments demonstrate that Sequencer performs impressively well: Sequencer2D-L, with 54M parameters, realizes 84.6% top-1 accuracy on only ImageNet-1K. Not only that, we show that it has good transferability and the robust resolution adaptability on double resolution-band.

Supervised or weakly supervised methods for phrase localization (textual grounding) either rely on human annotations or some other supervised models, e.g., object detectors. Obtaining these annotations is labor-intensive and may be difficult to scale in practice. We propose to leverage recent advances in contrastive language-vision models, CLIP, pre-trained on image and caption pairs collected from the internet. In its original form, CLIP only outputs an image-level embedding without any spatial resolution. We adapt CLIP to generate high-resolution spatial feature maps. Importantly, we can extract feature maps from both ViT and ResNet CLIP model while maintaining the semantic properties of an image embedding. This provides a natural framework for phrase localization. Our method for phrase localization requires no human annotations or additional training. Extensive experiments show that our method outperforms existing no-training methods in zero-shot phrase localization, and in some cases, it even outperforms supervised methods. Code is available at https://github.com/pals-ttic/adapting-CLIP .

Contrastive language-image pre-training, CLIP for short, has gained increasing attention for its potential in various scenarios. In this paper, we propose EVA-CLIP, a series of models that significantly improve the efficiency and effectiveness of CLIP training. Our approach incorporates new techniques for representation learning, optimization, and augmentation, enabling EVA-CLIP to achieve superior performance compared to previous CLIP models with the same number of parameters but significantly smaller training costs. Notably, our largest 5.0B-parameter EVA-02-CLIP-E/14+ with only 9 billion seen samples achieves 82.0 zero-shot top-1 accuracy on ImageNet-1K val. A smaller EVA-02-CLIP-L/14+ with only 430 million parameters and 6 billion seen samples achieves 80.4 zero-shot top-1 accuracy on ImageNet-1K val. To facilitate open access and open research, we release the complete suite of EVA-CLIP to the community at https://github.com/baaivision/EVA/tree/master/EVA-CLIP.

Vision Transformers (ViTs) take all the image patches as tokens and construct multi-head self-attention (MHSA) among them. Complete leverage of these image tokens brings redundant computations since not all the tokens are attentive in MHSA. Examples include that tokens containing semantically meaningless or distractive image backgrounds do not positively contribute to the ViT predictions. In this work, we propose to reorganize image tokens during the feed-forward process of ViT models, which is integrated into ViT during training. For each forward inference, we identify the attentive image tokens between MHSA and FFN (i.e., feed-forward network) modules, which is guided by the corresponding class token attention. Then, we reorganize image tokens by preserving attentive image tokens and fusing inattentive ones to expedite subsequent MHSA and FFN computations. To this end, our method EViT improves ViTs from two perspectives. First, under the same amount of input image tokens, our method reduces MHSA and FFN computation for efficient inference. For instance, the inference speed of DeiT-S is increased by 50% while its recognition accuracy is decreased by only 0.3% for ImageNet classification. Second, by maintaining the same computational cost, our method empowers ViTs to take more image tokens as input for recognition accuracy improvement, where the image tokens are from higher resolution images. An example is that we improve the recognition accuracy of DeiT-S by 1% for ImageNet classification at the same computational cost of a vanilla DeiT-S. Meanwhile, our method does not introduce more parameters to ViTs. Experiments on the standard benchmarks show the effectiveness of our method. The code is available at https://github.com/youweiliang/evit

Contrastive Language-Image Pre-training (CLIP) has become a foundation model and has been applied to various vision and multimodal tasks. However, recent works indicate that CLIP falls short in distinguishing detailed differences in images and shows suboptimal performance on dense-prediction and vision-centric multimodal tasks. Therefore, this work focuses on improving existing CLIP models, aiming to capture as many visual details in images as possible. We find that a specific type of generative models, unCLIP, provides a suitable framework for achieving our goal. Specifically, unCLIP trains an image generator conditioned on the CLIP image embedding. In other words, it inverts the CLIP image encoder. Compared to discriminative models like CLIP, generative models are better at capturing image details because they are trained to learn the data distribution of images. Additionally, the conditional input space of unCLIP aligns with CLIP's original image-text embedding space. Therefore, we propose to invert unCLIP (dubbed un^2CLIP) to improve the CLIP model. In this way, the improved image encoder can gain unCLIP's visual detail capturing ability while preserving its alignment with the original text encoder simultaneously. We evaluate our improved CLIP across various tasks to which CLIP has been applied, including the challenging MMVP-VLM benchmark, the dense-prediction open-vocabulary segmentation task, and multimodal large language model tasks. Experiments show that un^2CLIP significantly improves the original CLIP and previous CLIP improvement methods. Code and models will be available at https://github.com/LiYinqi/un2CLIP.

Image segmentation remains a challenging task in computer vision, demanding robust mask generation and precise classification. Recent mask-based approaches yield high-quality masks by capturing global context. However, accurately classifying these masks, especially in the presence of ambiguous boundaries and imbalanced class distributions, remains an open challenge. In this work, we introduce ViT-P, a novel two-stage segmentation framework that decouples mask generation from classification. The first stage employs a proposal generator to produce class-agnostic mask proposals, while the second stage utilizes a point-based classification model built on the Vision Transformer (ViT) to refine predictions by focusing on mask central points. ViT-P serves as a pre-training-free adapter, allowing the integration of various pre-trained vision transformers without modifying their architecture, ensuring adaptability to dense prediction tasks. Furthermore, we demonstrate that coarse and bounding box annotations can effectively enhance classification without requiring additional training on fine annotation datasets, reducing annotation costs while maintaining strong performance. Extensive experiments across COCO, ADE20K, and Cityscapes datasets validate the effectiveness of ViT-P, achieving state-of-the-art results with 54.0 PQ on ADE20K panoptic segmentation, 87.4 mIoU on Cityscapes semantic segmentation, and 63.6 mIoU on ADE20K semantic segmentation. The code and pretrained models are available at: https://github.com/sajjad-sh33/ViT-P}{https://github.com/sajjad-sh33/ViT-P.

We investigate the CLIP image encoder by analyzing how individual model components affect the final representation. We decompose the image representation as a sum across individual image patches, model layers, and attention heads, and use CLIP's text representation to interpret the summands. Interpreting the attention heads, we characterize each head's role by automatically finding text representations that span its output space, which reveals property-specific roles for many heads (e.g. location or shape). Next, interpreting the image patches, we uncover an emergent spatial localization within CLIP. Finally, we use this understanding to remove spurious features from CLIP and to create a strong zero-shot image segmenter. Our results indicate that a scalable understanding of transformer models is attainable and can be used to repair and improve models.

Deep learning models are increasingly deployed on resource-constrained edge devices for real-time data analytics. In recent years, Vision Transformer models and their variants have demonstrated outstanding performance across various computer vision tasks. However, their high computational demands and inference latency pose significant challenges for model deployment on resource-constraint edge devices. To address this issue, we propose a novel Vision Transformer splitting framework, ED-ViT, designed to execute complex models across multiple edge devices efficiently. Specifically, we partition Vision Transformer models into several sub-models, where each sub-model is tailored to handle a specific subset of data classes. To further minimize computation overhead and inference latency, we introduce a class-wise pruning technique that reduces the size of each sub-model. We conduct extensive experiments on five datasets with three model structures, demonstrating that our approach significantly reduces inference latency on edge devices and achieves a model size reduction of up to 28.9 times and 34.1 times, respectively, while maintaining test accuracy comparable to the original Vision Transformer. Additionally, we compare ED-ViT with two state-of-the-art methods that deploy CNN and SNN models on edge devices, evaluating accuracy, inference time, and overall model size. Our comprehensive evaluation underscores the effectiveness of the proposed ED-ViT framework.

High-resolution images offer more information about scenes that can improve model accuracy. However, the dominant model architecture in computer vision, the vision transformer (ViT), cannot effectively leverage larger images without finetuning -- ViTs poorly extrapolate to more patches at test time, although transformers offer sequence length flexibility. We attribute this shortcoming to the current patch position encoding methods, which create a distribution shift when extrapolating. We propose a drop-in replacement for the position encoding of plain ViTs that restricts attention heads to fixed fields of view, pointed in different directions, using 2D attention masks. Our novel method, called LookHere, provides translation-equivariance, ensures attention head diversity, and limits the distribution shift that attention heads face when extrapolating. We demonstrate that LookHere improves performance on classification (avg. 1.6%), against adversarial attack (avg. 5.4%), and decreases calibration error (avg. 1.5%) -- on ImageNet without extrapolation. With extrapolation, LookHere outperforms the current SoTA position encoding method, 2D-RoPE, by 21.7% on ImageNet when trained at 224^2 px and tested at 1024^2 px. Additionally, we release a high-resolution test set to improve the evaluation of high-resolution image classifiers, called ImageNet-HR.

We propose a new method for exemplar-free class incremental training of ViTs. The main challenge of exemplar-free continual learning is maintaining plasticity of the learner without causing catastrophic forgetting of previously learned tasks. This is often achieved via exemplar replay which can help recalibrate previous task classifiers to the feature drift which occurs when learning new tasks. Exemplar replay, however, comes at the cost of retaining samples from previous tasks which for many applications may not be possible. To address the problem of continual ViT training, we first propose gated class-attention to minimize the drift in the final ViT transformer block. This mask-based gating is applied to class-attention mechanism of the last transformer block and strongly regulates the weights crucial for previous tasks. Importantly, gated class-attention does not require the task-ID during inference, which distinguishes it from other parameter isolation methods. Secondly, we propose a new method of feature drift compensation that accommodates feature drift in the backbone when learning new tasks. The combination of gated class-attention and cascaded feature drift compensation allows for plasticity towards new tasks while limiting forgetting of previous ones. Extensive experiments performed on CIFAR-100, Tiny-ImageNet and ImageNet100 demonstrate that our exemplar-free method obtains competitive results when compared to rehearsal based ViT methods.

Joint Energy Models (JEMs), while drawing significant research attention, have not been successfully scaled to real-world, high-resolution datasets. We present EB-CLIP, a novel approach extending JEMs to the multimodal vision-language domain using CLIP, integrating both generative and discriminative objectives. For the generative objective, we introduce an image-text joint-energy function based on Cosine similarity in the CLIP space, training CLIP to assign low energy to real image-caption pairs and high energy otherwise. For the discriminative objective, we employ contrastive adversarial loss, extending the adversarial training objective to the multimodal domain. EB-CLIP not only generates realistic images from text but also achieves competitive results on the compositionality benchmark, outperforming leading methods with fewer parameters. Additionally, we demonstrate the superior guidance capability of EB-CLIP by enhancing CLIP-based generative frameworks and converting unconditional diffusion models to text-based ones. Lastly, we show that EB-CLIP can serve as a more robust evaluation metric for text-to-image generative tasks than CLIP.

Contrastive pretrained large Vision-Language Models (VLMs) like CLIP have revolutionized visual representation learning by providing good performance on downstream datasets. VLMs are 0-shot adapted to a downstream dataset by designing prompts that are relevant to the dataset. Such prompt engineering makes use of domain expertise and a validation dataset. Meanwhile, recent developments in generative pretrained models like GPT-4 mean they can be used as advanced internet search tools. They can also be manipulated to provide visual information in any structure. In this work, we show that GPT-4 can be used to generate text that is visually descriptive and how this can be used to adapt CLIP to downstream tasks. We show considerable improvements in 0-shot transfer accuracy on specialized fine-grained datasets like EuroSAT (~7%), DTD (~7%), SUN397 (~4.6%), and CUB (~3.3%) when compared to CLIP's default prompt. We also design a simple few-shot adapter that learns to choose the best possible sentences to construct generalizable classifiers that outperform the recently proposed CoCoOP by ~2% on average and by over 4% on 4 specialized fine-grained datasets. We will release the code, prompts, and auxiliary text dataset upon acceptance.

Vision Transformers (ViTs) have revolutionized computer vision by leveraging self-attention to model long-range dependencies. However, ViTs face challenges such as high computational costs due to the quadratic scaling of self-attention and the requirement of a large amount of training data. To address these limitations, we propose the Efficient Convolutional Vision Transformer (ECViT), a hybrid architecture that effectively combines the strengths of CNNs and Transformers. ECViT introduces inductive biases such as locality and translation invariance, inherent to Convolutional Neural Networks (CNNs) into the Transformer framework by extracting patches from low-level features and enhancing the encoder with convolutional operations. Additionally, it incorporates local-attention and a pyramid structure to enable efficient multi-scale feature extraction and representation. Experimental results demonstrate that ECViT achieves an optimal balance between performance and efficiency, outperforming state-of-the-art models on various image classification tasks while maintaining low computational and storage requirements. ECViT offers an ideal solution for applications that prioritize high efficiency without compromising performance.

Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this note, we show that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient. With simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ~1 day on a single 8-A100 node. We hope this can make state-of-the-art LMM research more accessible. Code and model will be publicly available.

Vision Transformers (ViTs) have shown remarkable performance and scalability across various computer vision tasks. To apply single-scale ViTs to image segmentation, existing methods adopt a convolutional adapter to generate multi-scale features, a pixel decoder to fuse these features, and a Transformer decoder that uses the fused features to make predictions. In this paper, we show that the inductive biases introduced by these task-specific components can instead be learned by the ViT itself, given sufficiently large models and extensive pre-training. Based on these findings, we introduce the Encoder-only Mask Transformer (EoMT), which repurposes the plain ViT architecture to conduct image segmentation. With large-scale models and pre-training, EoMT obtains a segmentation accuracy similar to state-of-the-art models that use task-specific components. At the same time, EoMT is significantly faster than these methods due to its architectural simplicity, e.g., up to 4x faster with ViT-L. Across a range of model sizes, EoMT demonstrates an optimal balance between segmentation accuracy and prediction speed, suggesting that compute resources are better spent on scaling the ViT itself rather than adding architectural complexity. Code: https://www.tue-mps.org/eomt/.

Vision Transformers (ViTs) have recently taken computer vision by storm. However, the softmax attention underlying ViTs comes with a quadratic complexity in time and memory, hindering the application of ViTs to high-resolution images. We revisit the attention design and propose a linear attention method to address the limitation, which doesn't sacrifice ViT's core advantage of capturing global representation like existing methods (e.g. local window attention of Swin). We further investigate the key difference between linear attention and softmax attention. Our empirical results suggest that linear attention lacks a fundamental property of concentrating the distribution of the attention matrix. Inspired by this observation, we introduce a local concentration module to enhance linear attention. By incorporating enhanced linear global attention and local window attention, we propose a new ViT architecture, dubbed L^2ViT. Notably, L^2ViT can effectively capture both global interactions and local representations while enjoying linear computational complexity. Extensive experiments demonstrate the strong performance of L^2ViT. On image classification, L^2ViT achieves 84.4% Top-1 accuracy on ImageNet-1K without any extra training data or label. By further pre-training on ImageNet-22k, it attains 87.0% when fine-tuned with resolution 384^2. For downstream tasks, L^2ViT delivers favorable performance as a backbone on object detection as well as semantic segmentation.

Vision Transformers (ViTs) achieve strong performance in image classification but incur high computational costs from processing all image tokens. To reduce inference costs in large ViTs without compromising accuracy, we propose TinyDrop, a training-free token dropping framework guided by a lightweight vision model. The guidance model estimates the importance of tokens while performing inference, thereby selectively discarding low-importance tokens if large vit models need to perform attention calculations. The framework operates plug-and-play, requires no architectural modifications, and is compatible with diverse ViT architectures. Evaluations on standard image classification benchmarks demonstrate that our framework reduces FLOPs by up to 80% for ViTs with minimal accuracy degradation, highlighting its generalization capability and practical utility for efficient ViT-based classification.

Vision language models (VLMs) have seen growing adoption in recent years, but many still struggle with basic spatial reasoning errors. We hypothesize that this is due to VLMs adopting pre-trained vision backbones, specifically vision transformers (ViTs) trained with image-level supervision and minimal inductive biases. Such models may fail to encode the class contents at each position in the image, and our goal is to resolve this by ensuring that the vision backbone effectively captures both local and global image semantics. Our main insight is that we do not require new supervision to learn this capability -- pre-trained models contain significant knowledge of local semantics that we can extract and use for scalable self-supervision. We propose a new efficient post-training stage for ViTs called locality alignment and a novel fine-tuning procedure called MaskEmbed that uses a masked reconstruction loss to learn semantic contributions for each image patch. We first evaluate locality alignment with a vision-only benchmark, finding that it improves a model's performance at a patch-level semantic segmentation task, especially for strong backbones trained with image-caption pairs (e.g., CLIP and SigLIP). We then train a series of VLMs with and without locality alignment, and show that locality-aligned backbones improve performance across a range of benchmarks, particularly ones that involve spatial understanding (e.g., RefCOCO, OCID-Ref, TallyQA, VSR, AI2D). Overall, we demonstrate that we can efficiently learn local semantic extraction via a locality alignment stage, and that this procedure complements existing VLM training recipes that use off-the-shelf vision backbones.