CyberDJ commited on
Commit
f6cb6cc
Β·
verified Β·
1 Parent(s): 354ba39

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +88 -0
README.md CHANGED
@@ -4,8 +4,96 @@ license: mit
4
  # Datacard
5
  This is the official fine-tuning dataset provided by VLABench, with 500 episodes each task. The current version includes 10 primitive tasks.
6
 
 
 
 
 
 
7
  ## Uses
8
  Download all archive files and use the following command to extract:
9
  ```sh
10
  cat rdt_data.tar.gz.* | tar -xzvf -
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11
  ```
 
4
  # Datacard
5
  This is the official fine-tuning dataset provided by VLABench, with 500 episodes each task. The current version includes 10 primitive tasks.
6
 
7
+ ## Source
8
+ - Project Page: [https://vlabench.github.io/](https://vlabench.github.io/)
9
+ - Arxiv Paper: [https://arxiv.org/abs/2412.18194](https://arxiv.org/abs/2412.18194)
10
+ - Code: [https://github.com/OpenMOSS/VLABench](https://github.com/OpenMOSS/VLABench)
11
+
12
  ## Uses
13
  Download all archive files and use the following command to extract:
14
  ```sh
15
  cat rdt_data.tar.gz.* | tar -xzvf -
16
+ ```
17
+
18
+ In the resulting `VLABench_release` folder, there will be two folders: `primitive`(release now) and `composite`(under management).
19
+ In `primitive` folder, there are ten sub-folders and the dataset HDF5 files can be listed as:
20
+
21
+ ```
22
+ VLABench_release
23
+ └── primitive
24
+ β”œβ”€β”€ add_condiment
25
+ └── episode_0.hdf5
26
+ ...
27
+ β”œβ”€β”€ insert_flower
28
+ └── episode_0.hdf5
29
+ ...
30
+ β”œβ”€β”€ select_book
31
+ └── ...
32
+ β”œβ”€β”€ select_chemistry_tube
33
+ └── ...
34
+ β”œβ”€β”€ select_drink
35
+ └── ...
36
+ β”œβ”€β”€ select_fruit
37
+ └── ...
38
+ β”œβ”€β”€ select_mahjong
39
+ └── ...
40
+ β”œβ”€β”€ select_painting
41
+ └── ...
42
+ β”œβ”€β”€ select_poker
43
+ └── ...
44
+ └── select_toy
45
+ └── ...
46
+ ```
47
+
48
+ An example of the single episode data is:
49
+ ```
50
+ data
51
+ 2025-02-23 20:46:40
52
+ instruction
53
+ (1,) |S38
54
+ ['Please put the striped_10 in any hole.']
55
+ meta_info
56
+ entities
57
+ (6,) |S15
58
+ ['billiards_table', 'striped_10', 'striped_14', 'striped_11', 'striped_12', 'solid_1']
59
+ episode_config
60
+ () | S1879
61
+ target_entity
62
+ (1,) |S10
63
+ ['striped_10']
64
+ observation
65
+ depth
66
+ (212, 4, 480, 480) float32
67
+ ee_state
68
+ (212, 8) float32
69
+ point_cloud_colors
70
+ (212, 11905, 3) float32
71
+ point_cloud_points
72
+ (212, 11905, 3) float32
73
+ q_acceleration
74
+ (212, 7, 1) float32
75
+ q_state
76
+ (212, 7, 1) float32
77
+ q_velocity
78
+ (212, 7, 1) float32
79
+ rgb
80
+ (212, 4, 480, 480, 3) uint8
81
+ robot_mask
82
+ (212, 4, 480, 480) float32
83
+ trajectory
84
+ (212, 8) float32
85
+ ```
86
+
87
+ ## Citation
88
+ If you find our work helps,please cite us:
89
+ ```
90
+ @misc{zhang2024vlabench,
91
+ title={VLABench: A Large-Scale Benchmark for Language-Conditioned Robotics Manipulation with Long-Horizon Reasoning Tasks},
92
+ author={Shiduo Zhang and Zhe Xu and Peiju Liu and Xiaopeng Yu and Yuan Li and Qinghui Gao and Zhaoye Fei and Zhangyue Yin and Zuxuan Wu and Yu-Gang Jiang and Xipeng Qiu},
93
+ year={2024},
94
+ eprint={2412.18194},
95
+ archivePrefix={arXiv},
96
+ primaryClass={cs.RO},
97
+ url={https://arxiv.org/abs/2412.18194},
98
+ }
99
  ```