Skip to content

Dataset · facebook

First-person hand tracking dataset (Facebook / Meta)

facebook/hand_tracking_challenge_umetrack

Large image and text dataset for the Multiview Egocentric Hand Tracking Challenge 2024.

What's inside

Facebook dataset for multiview egocentric hand tracking tasks. Includes image and text data with a total volume of 100 thousand to 1 million examples. To work with the dataset, the authors suggest using the specialized hand_tracking_toolkit tool.

What you can do with it

  • Training computer vision models for 3D tracking and hand pose estimation in VR/AR headsets
  • Training multiview first-person gesture recognition algorithms
  • Testing models on data from the Multiview Egocentric Hand Tracking Challenge 2024
Size
100K–1M rows
Downloads this month
2.4K
Likes
3

How to load it

from datasets import load_dataset

ds = load_dataset("facebook/hand_tracking_challenge_umetrack")