RevoTac-OXT / metadata.json
DavidLXu's picture
Group all task Zarr archives under the RevoTac-OXT dataset directory
0104b9b verified
Raw History Blame Contribute Delete
1.39 kB
{
"name": "RevoTac-OXT",
"url": [
"https://huggingface.co/datasets/BrainCoTech/RevoTac-OXT"
],
"version": "0.1.0",
"description": "Ten scripted simulated tactile manipulation episodes from RevoBench: bimanual Revo3 dexterous hands; 32272 frames at 60 Hz. Rendered Taxim RGB, numeric indentation depth, HydroShear marker displacements, normalized pressure response, and target-specific PhysX pad contact wrench. Simulated data; development trajectories, one episode per task. Stable long holds removed offline without interpolation; original capture frame indices and times retained under edit_provenance. Not physics re-executed. Contributing institutions: HKUST(GZ), BrainCo, Georgia Tech.",
"type": "dexhand",
"sensor_image": [
"RevoBenchSimTaxim"
],
"sensor_array": [
"RevoBenchSimDepth"
],
"sensor_state": [
"RevoBenchSimHydroShear",
"RevoBenchSimPhysXContactWrench",
"RevoBenchSimPressure"
],
"task": 10,
"trajectory": 10,
"frame": 32272,
"label": [
"ego"
],
"contact": [
"davidxulixin@gmail.com"
],
"reference": {
"paper": "RevoSim: Scalable Multimodal Tactile Simulation for Dexterous Manipulation",
"website": "https://braincotech.github.io/RevoSim-web/",
"authors": "Ke Zhang, Lixin Xu, Ziyi Wang, Xiyue Dong, Jie Tan, Lianghu Shao, Xianpeng Wang, Chuanyu Li, Renjing Xu."
},
"code_trigger": ""
}