Dexterous Grasp Transformer
iSEE Laboratory
A transformer-based method for generating dexterous grasps from object geometry.
What you can explore
Inspect the training pipeline, generated grasps and evaluation against the underlying grasp data.
Before you use it
Uses preprocessed DexGraspNet inputs and specific CUDA/PyTorch dependencies. Object meshes and grasp data require their own terms; generated grasps are not physical performance guarantees.
Original contributors
Guo-Hao Xu, Yi-Lin Wei, Dian Zheng, Xiao-Ming Wu and Wei-Shi Zheng
License / access: Code: MIT · dataset and mesh terms are separate
Editorial reference · checked 2026-09-22