yeszhong2026cc-by-nc-sa-4.0

AFUN_train

Training dataset of 44,749 data points for affordance segmentation and 3D interaction-motion prediction, featuring RGB frames, depth maps, affordance masks, 3D motion trajectories, and language instructions.

Downloads68
Episodes44749

Why This Matters for Physical AI

AFUN provides a large-scale multimodal dataset for learning affordance understanding and 3D motion prediction, enabling foundation models to recognize functional object parts and predict how they interact across diverse robotic and human demonstrations.

Technical Profile

Modalities
rgbdepthlanguagesegmentation
Robot Embodiments
Franka Pandahumanoidmobile_manipulator
Environment
labsimulationindoor
Task Types
manipulationaffordance_predictionmotion_prediction
Episodes
44749
Data Format
custom
Annotation Types
language_instructionssegmentation3d_motion_trajectories
License
cc-by-nc-sa-4.0
Part of the AFUN family

Access

Need custom rgb data?

Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets