Stereo Feature Learning based on Attention and Geometry for Absolute Hand Pose Estimation in Egocentric Stereo Views

SEH

• Our SEH dataset is the first hand pose dataset for egocentric stereo vision with accurate GT 3D hand poses.
• Our dataset includes three types of interactions : tapping, drawing, and toruch gestures. These interactions were performed on three types of backgrounds: white, wood, and cluttered.

Dataset Download

1. Download Data.zip file (https://drive.google.com/file/d/1r4lkzssT4FlFAlKMvEV7EZ3LGJ7zxomJ/view?usp=sharing)
2. Extract Data.zip

Directory

Data

|-- all_data

|-- labels