ViTacFormer
ViTacFormer research team
Learning and deployment code for combining vision and tactile signals in dexterous manipulation.
What you can explore
Inspect training and inference with paired visual and tactile observations.
Before you use it
The documented setup uses two SharpaWave hands, tactile sensors, bimanual arms and active vision. Example data is linked separately; another hand or sensor needs its own integration and validation.
Original contributors
Liang Heng, Haoran Geng, Kaifeng Zhang, Pieter Abbeel and Jitendra Malik
License / access: Code: Apache 2.0 · verify separate example-data and hardware terms
Keep exploring
Sharpa Wave SDK
Libraries and API documentation for integrating Sharpa Wave into robotics applications.
Sparsh
Self-supervised encoders for vision-based touch, with downstream evaluation through TacBench.
Editorial reference · checked 2026-09-22