futurefantasy commited on
Commit
9843d61
·
verified ·
1 Parent(s): e7a621b

Finalize simplified release interface for README.md

Browse files
Files changed (1) hide show
  1. README.md +22 -14
README.md CHANGED
@@ -21,37 +21,45 @@ VLAC2 is a video-progress benchmark for long-horizon robot manipulation. This re
21
  - `benchmark_splits/test_nonexpert_unseen/video_progress_benchmark_file.json`
22
  - `scripts/extract_vlac2_release_raw_archives.sh`
23
  - `scripts/extract_vlac2_release_frames.py`
24
- - `scripts/vlac2_release_common.py`
25
  - `data/train_videos.tar`
26
  - `data/test_videos.tar`
27
 
28
  ## Usage
29
 
30
- 1. Unpack the benchmark video archives directly into `data/`.
31
 
32
  ```bash
33
- bash scripts/extract_vlac2_release_raw_archives.sh \
34
- /path/to/vlac2_release/data \
35
- /path/to/vlac2_release/data
36
  ```
37
 
38
- If both arguments are omitted, the script defaults to `data/` and extracts into `data/` under the release root.
39
- The script expects exactly the released benchmark archives: `test_videos.tar` and `train_videos.tar` (or their `.tar.zst` variants).
40
 
41
- 2. Extract benchmark frames from the unpacked raw videos.
 
 
 
 
 
 
 
 
 
 
 
 
42
 
43
  ```bash
44
- python scripts/extract_vlac2_release_frames.py --release-root /path/to/vlac2_release
45
  ```
46
 
47
- Here `--release-root` should be the top-level release directory that contains `README.md`, `benchmark_splits/`, `data/`, and `scripts/`.
48
- If `--output-root` is not specified, extracted frames are written to `/path/to/vlac2_release/_extracted_frames/`, which is created automatically by the script.
49
 
50
- 3. Use the benchmark files for training or evaluation. Frame paths in the JSONs use the prefix `__VLAC2_FRAMES_ROOT__/...`. Replace `__VLAC2_FRAMES_ROOT__` with the absolute path of the extracted-frame directory. For example, if extracted frames are written to `/data/VLAC2-Benchmark/_extracted_frames`, then `__VLAC2_FRAMES_ROOT__/ARX-data/...` becomes `/data/VLAC2-Benchmark/_extracted_frames/ARX-data/...`. If you choose a custom `--output-root`, substitute that directory instead.
 
51
 
52
  ## Notes
53
 
54
- - The benchmark JSON files are ready to use; users do not need to rebuild them.
55
  - A video decoder backend is required for frame extraction: `opencv-python` or `imageio`.
56
  - This repo only contains the raw videos required by the released benchmark splits. The full raw-data release is distributed separately.
57
- - The raw-video subset and the extracted-frame cache are intentionally separated so that users can regenerate frames locally under their own storage layout.
 
21
  - `benchmark_splits/test_nonexpert_unseen/video_progress_benchmark_file.json`
22
  - `scripts/extract_vlac2_release_raw_archives.sh`
23
  - `scripts/extract_vlac2_release_frames.py`
 
24
  - `data/train_videos.tar`
25
  - `data/test_videos.tar`
26
 
27
  ## Usage
28
 
29
+ 1. Unpack the video archives.
30
 
31
  ```bash
32
+ bash scripts/extract_vlac2_release_raw_archives.sh
 
 
33
  ```
34
 
35
+ This extracts `data/test_videos.tar` and `data/train_videos.tar` into `data/`.
36
+ To extract the raw videos elsewhere, pass a target directory:
37
 
38
+ ```bash
39
+ bash scripts/extract_vlac2_release_raw_archives.sh /path/to/data
40
+ ```
41
+
42
+ 2. Extract frames.
43
+
44
+ ```bash
45
+ python scripts/extract_vlac2_release_frames.py --data-root data
46
+ ```
47
+
48
+ This writes extracted frames to `_extracted_frames/` under the release directory.
49
+ Only the benchmark main view is extracted.
50
+ If the raw videos were extracted elsewhere:
51
 
52
  ```bash
53
+ python scripts/extract_vlac2_release_frames.py --data-root /path/to/data
54
  ```
55
 
56
+ 3. Use the benchmark splits.
 
57
 
58
+ Frame paths in the JSON files use the prefix `__VLAC2_FRAMES_ROOT__/...`.
59
+ Replace `__VLAC2_FRAMES_ROOT__` with the absolute path of `_extracted_frames`.
60
 
61
  ## Notes
62
 
 
63
  - A video decoder backend is required for frame extraction: `opencv-python` or `imageio`.
64
  - This repo only contains the raw videos required by the released benchmark splits. The full raw-data release is distributed separately.
65
+ - The benchmark JSON files are ready to use. No benchmark reconstruction is required.