Open-AoE: An Open Egocentric Manipulation Dataset and Toolchain for Embodied Learning
Open-AoE pairs about 2,000 hours of smartphone-captured manipulation video with annotations and robot-learning tools.
The first release was collected in natural environments by more than 500 contributors using over 400 smartphones. It includes text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. The paper also describes processing tools for segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction, plus downstream support for visualization, retargeting, data conversion, and training recipes. HF Daily Papers' note
The first release was collected in natural environments by more than 500 contributors using over 400 smartphones. It includes text annotations, MANO-based hand poses, camera trajectories, and temporally localized atomic actions. The paper also describes processing tools for segmentation, semantic annotation, hand reconstruction, and camera trajectory reconstruction, plus downstream support for visualization, retargeting, data conversion, and training recipes. HF Daily Papers' note
score 6