Foundation-model-based task-oriented grasping from SUSTech: combines semantic (language) and geometric reasoning to propose task-appropriate 6-DoF grasps that generalise to novel objects and tasks.
Input
multi-modal
Output
6-DOF grasp pose
Foundation-model-based task-oriented grasping from SUSTech: combines semantic (language) and geometric reasoning to propose task-appropriate 6-DoF grasps that generalise to novel objects and tasks.
Save it, compare it, request a quote or record that you have used it — with an account.
FoundationGrasp is a task-oriented grasping (TOG) framework from the Robotics and Computer Vision Lab at Southern University of Science and Technology (SUSTech). It leverages the open-ended knowledge in foundation models to select grasps appropriate for the intended task (e.g. grasp a knife by the handle to cut), combining language and vision through the LaViA-TaskGrasp dataset. It extends the group's earlier GraspGPT work and generalises beyond training data to novel objects, categories, and task instructions. Code, data and appendix are published via the project page; the implementation lives in the GraspGPT public repository.
This product is part of the Birdwave Atlas community.
Contribute evaluations, share integrations and help improve compatibility data.
Discussions, evaluations, integrations and updates