| { | |
| "name": "ViTaMIn", | |
| "url": [ | |
| "https://huggingface.co/datasets/chuanyune/ViTaMIn-OXT" | |
| ], | |
| "version": "1.0.1", | |
| "description": "OXT-formatted derivative of the latest public ViTaMIn contact-rich manipulation dataset. It contains 6 single-hand UMI-style visuo-tactile tasks (747 trajectories, 361339 frames) with wrist RGB, left/right tactile images, wrist pose, and parallel-gripper opening width in metres. Larger gripper values mean more open, and the scalar is mapped to canonical FTP-1 hand/gripper slot 28. Data are processed at 30 FPS; timestamps are synthetic per-episode seconds because the public source Zarr archives do not contain a timestamp stream. Version 1.0.1 corrects all wrist and demonstration start/end rotations to the FTP-1 tool convention by right-multiplying each source rotation by Rz(+90 degrees); XYZ positions and all image/tactile/gripper streams are unchanged.", | |
| "type": "gripper", | |
| "sensor_image": [ | |
| "ViTaMIn" | |
| ], | |
| "sensor_array": [], | |
| "sensor_state": [], | |
| "task": 6, | |
| "trajectory": 747, | |
| "frame": 361339, | |
| "label": [ | |
| "umi" | |
| ], | |
| "contact": [ | |
| "chuanyu.ne79@gmail.com" | |
| ], | |
| "reference": { | |
| "paper": "ViTaMIn: Learning Contact-Rich Tasks Through Robot-Free Visuo-Tactile Manipulation Interface (arXiv:2504.06156)", | |
| "website": "https://chuanyune.github.io/ViTaMIn_page/", | |
| "authors": "Fangchen Liu, Chuanyu Li, Yihua Qin, Jing Xu, Pieter Abbeel, Rui Chen" | |
| }, | |
| "code_trigger": "FTP_1_Source" | |
| } | |