The UniVTAC Benchmark dataset provides simulation data for tactile-based robotic manipulation tasks.
This dataset contains 100 episodes per task, totaling 800 episodes across 8 diverse manipulation tasks.
UniVTAC Benchmark currently includes the following manipulation tasks, all featuring tactile sensing:
| Task | Module | Description |
|---|---|---|
| Collect | collect | Collect contact-rich tactile data for pretraining |
| Lift Bottle | lift_bottle | Grasp and lift a bottle off a surface near a wall |
| Lift Can | lift_can | Grasp and lift a cylindrical can |
| Insert HDMI | insert_HDMI | Insert an HDMI connector into a port |
| Insert Hole | insert_hole | Precision peg-in-hole insertion |
| Insert Tube | insert_tube | Insert a tube into a fixture |
| Pull Out Key | pull_out_key | Extract a key from a lock |
| Put Bottle in Shelf | put_bottle_in_shelf | Place a bottle onto a shelf |
| Grasp & Classify | grasp_classify | Grasp an object and classify it by tactile feedback |
For detailed instructions on data loading, environment setup, and benchmarking protocols, please visit our:
π official website
π Github repo
68 commits
The UniVTAC Benchmark dataset provides simulation data for tactile-based robotic manipulation tasks.
This dataset contains 100 episodes per task, totaling 800 episodes across 8 diverse manipulation tasks.
UniVTAC Benchmark currently includes the following manipulation tasks, all featuring tactile sensing:
| Task | Module | Description |
|---|---|---|
| Collect | collect | Collect contact-rich tactile data for pretraining |
| Lift Bottle | lift_bottle | Grasp and lift a bottle off a surface near a wall |
| Lift Can | lift_can | Grasp and lift a cylindrical can |
| Insert HDMI | insert_HDMI | Insert an HDMI connector into a port |
| Insert Hole | insert_hole | Precision peg-in-hole insertion |
| Insert Tube | insert_tube | Insert a tube into a fixture |
| Pull Out Key | pull_out_key | Extract a key from a lock |
| Put Bottle in Shelf | put_bottle_in_shelf | Place a bottle onto a shelf |
| Grasp & Classify | grasp_classify | Grasp an object and classify it by tactile feedback |
For detailed instructions on data loading, environment setup, and benchmarking protocols, please visit our:
π official website
π Github repo
68 commits