FVL-Repo commited on
Commit
556fe04
·
verified ·
1 Parent(s): caa9571

Update dataset card for LeRobot release

Browse files
Files changed (1) hide show
  1. README.md +7 -3
README.md CHANGED
@@ -1,5 +1,6 @@
1
  ---
2
  pretty_name: OpenViTac Benchmark Dataset
 
3
  tags:
4
  - robotics
5
  - visuo-tactile manipulation
@@ -7,15 +8,15 @@ tags:
7
 
8
  # OpenViTac Benchmark Dataset
9
 
10
- OpenViTac is a visuo-tactile robot manipulation benchmark covering physical-property perception, fragile-object handling, contact-rich manipulation, and precise insertion. This release contains **1,300 simulated trajectories** in HDF5 format across 11 tasks.
11
 
12
  For the benchmark and data-collection details, see the [paper](https://arxiv.org/abs/2610.10384v1) and [code repository](https://github.com/FVL-Repo/OpenViTac).
13
 
14
  ## Simulation tasks
15
 
16
- Each task is stored under `Simulation/GelSight_mini/<task>/`. Some tasks have subdirectories for object properties or variants.
17
 
18
- | Capability | Task directory | Task summary | Trajectories |
19
  | --- | --- | --- | ---: |
20
  | Physical-property perception | `weight_classify` | Identify whether a block is light or heavy through interaction and place it on the matching target. | 100 |
21
  | Physical-property perception | `hardness_classify` | Identify whether a block is soft or hard through contact and place it on the matching target. | 100 |
@@ -29,4 +30,7 @@ Each task is stored under `Simulation/GelSight_mini/<task>/`. Some tasks have su
29
  | Precision manipulation | `insert_block_v1` | Align and insert a designated block into the matching opening. | 100 |
30
  | Precision manipulation | `insert_block_v2` | Select the instructed object, then insert it into the matching opening. | 300 |
31
 
 
 
32
  ## Real-world tasks
 
 
1
  ---
2
  pretty_name: OpenViTac Benchmark Dataset
3
+ license: apache-2.0
4
  tags:
5
  - robotics
6
  - visuo-tactile manipulation
 
8
 
9
  # OpenViTac Benchmark Dataset
10
 
11
+ OpenViTac is a visuo-tactile robot manipulation benchmark covering physical-property perception, fragile-object handling, contact-rich manipulation, and precise insertion. This release contains **1,500 simulated episodes in LeRobot v2.1 format** across 11 tasks: 1,300 with GelSight mini and 100 each with Neote and Xense for `insert_USB`.
12
 
13
  For the benchmark and data-collection details, see the [paper](https://arxiv.org/abs/2610.10384v1) and [code repository](https://github.com/FVL-Repo/OpenViTac).
14
 
15
  ## Simulation tasks
16
 
17
+ Each `Simulation/<sensor>/<task>/` directory is a separate LeRobot dataset. Its `meta/` directory contains episode metadata, `data/chunk-000/` contains one Parquet file per episode, and `videos/chunk-000/` contains the corresponding camera and tactile videos. GelSight mini data is under `Simulation/GelSight_mini/`.
18
 
19
+ | Capability | Task directory | Task summary | Episodes |
20
  | --- | --- | --- | ---: |
21
  | Physical-property perception | `weight_classify` | Identify whether a block is light or heavy through interaction and place it on the matching target. | 100 |
22
  | Physical-property perception | `hardness_classify` | Identify whether a block is soft or hard through contact and place it on the matching target. | 100 |
 
30
  | Precision manipulation | `insert_block_v1` | Align and insert a designated block into the matching opening. | 100 |
31
  | Precision manipulation | `insert_block_v2` | Select the instructed object, then insert it into the matching opening. | 300 |
32
 
33
+ The `insert_USB` task also includes Neote and Xense tactile sensor data, with 100 episodes each in `Simulation/Neote/insert_USB/` and `Simulation/Xense/insert_USB/`.
34
+
35
  ## Real-world tasks
36
+