Paper Website

Yufeng Tian*, Shuiqi Cheng*, Tianming Wei, Tianxing Zhou, Yuanhang Zhang, Zixian Liu, Qianwei Han, Zhecheng Yuan, Huazhe Xu.

* Equal contribution.

We propose Visual Tactile Soft Fusion Contrastive Learning (ViTaS), a novel visuo-tactile representation learning framework for visuomotor learning.

Installation

conda env create -f environment.yml
cd third_party && pip install -r requirements.txt && cd ..

Training ViTaS

bash train.sh

For more tasks and algorithms, please refer to train.sh for more details.

Acknowledgement

Our code is generally built upon M3L and tactile_envs. Tactile sensors used in real-world experiments are built with the help of 3D-ViTaC. We thank all these authors for their nicely open sourced code and their great contributions to the community.