---
license: mit
language:
- en
tags:
- computer-vision
- robotics
- robot-learning
---

# Dataset Card for 2HANDS
<!-- ## Dataset Details -->
2HANDS is the 2-Handed Affordance + Narration DataSet, consisting of a large number of unimanual and bimanual object affordance segmentation masks and task narrations as affordance class-labels.  

- **Project Site** https://sites.google.com/view/2handedafforder (ICCV 2025)
- **Paper:** https://arxiv.org/abs/2503.09320
- **Repository:** https://github.com/pearl-robot-lab/2HandedAfforder

Egocentric images and narrations/verb classes are derived from the EPIC-KITCHENS dataset and EPIC-VISOR annotations [1, 2].  
[1] Damen, D. et al. (2018). Scaling egocentric vision: The epic-kitchens dataset. ECCV 2018  
[2] Darkhalil, A. et al. (2022). Epic-kitchens visor benchmark: Video segmentations and object relations. NeurIPS 2022  

## Citation
You may cite our work as:  
Heidinger, M.\*, Jauhri, S.\*, Prasad, V., & Chalvatzaki, G. (2025). 2handedafforder: Learning precise actionable bimanual affordances from human videos. ICCV 2025

**BibTeX:**  
@misc{heidinger20252handedafforderlearningpreciseactionable,
      title={2HandedAfforder: Learning Precise Actionable Bimanual Affordances from Human Videos},
      author={Marvin Heidinger and Snehal Jauhri and Vignesh Prasad and Georgia Chalvatzaki},
      year={2025},
      eprint={2503.09320},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
}