Skip to content

Dataset · WANG-XIXI

Benchmark for 3D hand reconstruction on egocentric video

WANG-XIXI/EgoHandBench

A unified set of seven datasets in a single format for evaluating 3D hand recovery models.

What's inside

A dataset containing between 1k and 10k clips and frames (including sources such as ARCTIC), created to test 3D hand reconstruction models. The authors standardized seven open datasets into a single format, filtered out frames with poor annotations, and established train/test splits. Ready-to-use baselines and an evaluation script are also included on GitHub.

What you can do with it

  • Comparing and testing hand 3D reconstruction models on egocentric video
  • Training hand tracking systems for VR/AR and robotics on cleaned data
Size
1K–10K rows
Downloads this month
3.3K
Likes
3

How to load it

from datasets import load_dataset

ds = load_dataset("WANG-XIXI/EgoHandBench")