ReachingField

This repository is the official implementation of the ReachingField model from our 3DV 2025 paper:

3D Whole-body Grasp Synthesis with Directional Controllability (https://gpaschalidis.github.io/cwgrasp/).

ReachingField is a probabilistic 3D vector field conditioned on a 3D object. Specifically, it defines a distribution over the directions from which a body's arm and hand can reach an object, without penetrating obstacles around the object. Below, we provide visualizations illustrating how ReachingField is formed for objects placed at various heights, ranging from low to high, on different receptacle objects.

Installation & Dependencies

Clone the repository using:

git clone [email protected]:gpaschalidis/ReachingField.git
cd ReachingField

Run the following commands:

conda create -n reachingfield python=3.9 -y
conda activate reachingfield
conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install -r requirements.txt

Installing Dataset

To try ReachingField you can use any data you want. However, for our paper we use the grasping objects from GRAB dataset and the ReplicaGrasp dataset.

• Download first the GRAB object mesh (.ply) files from the GRAB website.
• Download the ReplicaGrasp dataset from FLEX github page.
• Create a folder and paste both datasets inside it, so that they follow the structure below:

      data
       │
       ├── contact_meshes
       │    ├── airplane.ply
       │    └── ...
       │
       └── replicagrasp
            ├── dset_info.npz
            └── receptacles.npz

Testing ReachingField

To try ReachingField and generate visualizations for each of its building steps, use the following command:

python test_reachingfield.py --obj_rec_conf cup_receptacle_aabb_WC1_Top3frl_apartment_wall_cabinet_01_all_1 --grasp_type "right" --vis

Inside the file dset_info.npz you will find a dictionary containing all the configurations of the ReplicaGrasp dataset. Each dictionary's key represents a distinct configuration. You can experiment with any configuration by replacing cup_receptacle_aabb_WC1_Top3frl_apartment_wall_cabinet_01_all_1 in the command above with the corresponding key.

Citation

If you found this work influential or helpful for your research, please cite:

@inproceedings{paschalidis2025cwgrasp,
  title     = {{3D} {W}hole-Body Grasp Synthesis with Directional Controllability},
  author    = {Paschalidis, Georgios and Wilschut, Romana and Anti\'{c}, Dimitrije and Taheri, Omid and Tzionas, Dimitrios},
  booktitle = {{International Conference on 3D Vision (3DV)}},
  year      = {2025}
 }