Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,19 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
|
| 5 |
+
## FlowOcc3D (flow dataset)
|
| 6 |
+
|
| 7 |
+
Download [FlowOcc3D](https://drive.google.com/file/d/1Xo2A40bUYA1wwXxqi_y6sQqu5mRRkc-s/view?usp=sharing). Unzip it in `./data/nuscenes`.
|
| 8 |
+
|
| 9 |
+
Here we briefly introduce what the downloaded file looks like. We store the flow and index of the foreground voxels in the `.bin` file and the `_idx.bin` file.
|
| 10 |
+
```python
|
| 11 |
+
W, H, Z = 200, 200, 16
|
| 12 |
+
sample_idx = results['sample_idx'] # nuScenes sample token
|
| 13 |
+
data_path = os.path.join('./data/nuscenes', 'occ_flow_sparse_ext', sample_idx)
|
| 14 |
+
|
| 15 |
+
occ_flow = np.ones((W*H*Z, 2)) * pad_value # pad_value could be zero
|
| 16 |
+
sparse_flow = np.fromfile(data_path + '.bin', dtype=np.float16).reshape(-1, 3)[:, :2]
|
| 17 |
+
sparse_idx = np.fromfile(data_path + '_idx.bin', dtype=np.int32).reshape(-1)
|
| 18 |
+
occ_flow[sparse_idx] = sparse_flow
|
| 19 |
+
occ_flow = occ_flow.reshape(W, H, Z, 2)
|