CoToGrasp: Contact-Topology-Conditioned Dexterous Grasp Synthesis via Canonical Workspace Learning

Julien Mérand¹,    Boris Meden¹,    Liming Chen²,    Mathieu Grossard¹

¹Université Paris-Saclay, CEA-List    ²École Centrale de Lyon, CNRS, LIRIS, UMR5205, Institut Universitaire de France (IUF)

19th European Conference on Computer Vision (ECCV 2026)

Project Page   |   arXiv

CoToGrasp is a novel generative framework that synthesizes diverse and stable grasps strictly conditioned on specific contact topologies.

See https://github.com/CEA-LIST/CoToGrasp for full code and instructions.

Your directory tree should strictly follow this structure:

# Model Checkpoints
COTOGRASP                                           # Main folder
├── ...
├── logs                                            # ckpts folder
    ├── allegro_right_goag_dgcnn_types_2_0209
    └── shadowhand_goag_dgcnn_types_2_0128
└── ...

# Dataset Directory
COTOGRASP_DATA
├── handprints
├── pointclouds
    ├── dexgraspnet
    ├── multidex
        ├── contactdb
        └── ycb
├── urdf
    ├── objects
        ├── dexgraspnet
        ├── multidex
    └── robot
└── workspaces