File size: 3,194 Bytes
51632a9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1235883
 
 
 
 
 
 
 
 
51632a9
5d495c4
 
 
 
 
 
51632a9
 
 
 
 
 
 
 
 
 
 
5d495c4
4158e7f
be9e4c5
 
 
51632a9
 
 
 
 
 
 
 
be9e4c5
 
 
 
 
 
 
 
 
51632a9
 
 
 
 
 
 
 
 
 
 
 
 
 
1235883
51632a9
 
5d495c4
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
---
pretty_name: VLAC-Cut-FullData
tags:
- robotics
- video-progress
- benchmark
- embodied-ai
license: other
---

# VLAC-Cut-FullData

VLAC-Cut-FullData is the full-data release for VLAC-Cut. It provides the complete raw-data archive set, benchmark-style JSON files, and a lightweight frame-extraction workflow for reproducing evaluation on the released benchmark protocol.

## Contents

```text
benchmark_style_all/
  train/video_progress_benchmark_file.json
  test_expert_seen/video_progress_benchmark_file.json
  test_expert_unseen/video_progress_benchmark_file.json
  test_nonexpert_seen/video_progress_benchmark_file.json
  test_nonexpert_unseen/video_progress_benchmark_file.json
  all/video_progress_benchmark_file.json

scripts/
  unpack_data.sh
  extract_vlac2_release_frames.py

annotation_files_raw/
  0106/
  0109/
  0116/
  1231/
  dataset_ver/
  hfx_project-1/
  refine_data/

data/
  vlac2_release_data_part01_arx_group_a.tar.zst.part-*
  vlac2_release_data_part02_arx_group_b.tar.zst.part-*
  vlac2_release_data_part03_arx_group_c.tar.zst.part-*
  vlac2_release_data_part04_droid_group_a.tar.zst.part-*
  vlac2_release_data_part05_droid_group_b.tar.zst.part-*
  vlac2_release_data_part06_droid_group_c.tar.zst.part-*
  vlac2_release_data_part07_other_sources.tar.zst
```

## Usage

### 1. Unpack the raw-data archives

```bash
bash scripts/unpack_data.sh /path/to/data
```

Several large archives are distributed as multi-part files to satisfy Hub file-size limits. `unpack_data.sh` detects these parts automatically and streams them into a single extraction pass.

### 2. Extract frames

To extract frames only for one or more selected benchmark JSON files:

```bash
python scripts/extract_vlac2_release_frames.py \
  --data-root /path/to/data \
  --frames-root /path/to/data_extracted_frames \
  --benchmark-json benchmark_style_all/test_expert_seen/video_progress_benchmark_file.json
```

This decodes only the episodes referenced by the provided benchmark JSON file(s), rather than scanning the full raw-data release.

If `--benchmark-json` is omitted, the extractor scans all benchmark JSON files under `benchmark_style_all/`, collects all referenced records, and extracts the corresponding `main_path` videos:

```bash
python scripts/extract_vlac2_release_frames.py \
  --data-root /path/to/data \
  --frames-root /path/to/data_extracted_frames
```

### 3. Use the benchmark JSON files

Frame paths in the benchmark JSON files use the prefix:

```text
__VLAC2_FRAMES_ROOT__/
```

Replace `__VLAC2_FRAMES_ROOT__` with the absolute path to the extracted-frame directory.

## Notes

* This release contains the complete raw-data archive set used by the VLAC-Cut benchmark release workflow.
* `annotation_files_raw/` mirrors the original raw annotation JSON tree used during benchmark construction.
* The `benchmark_style_all/all` split is larger than the benchmark evaluation subsets and is rebuilt from the full annotation PKL source tree.
* The benchmark JSON files are pre-generated and do not need to be rebuilt for standard use.
* The large raw-data archives are distributed as split `.part-*` files in `data/`; these are unpacked transparently by the provided script.