Skip to content

Dataset · Voxel51

MPII Human Pose annotations on images

Voxel51/MPII_Human_Pose_Dataset

About 25K YouTube frames with joint annotations for over 40K people for pose and action estimation.

What's inside

A well-known benchmark for the human pose estimation task. It contains about 25,000 images extracted from YouTube videos, with joint coordinates annotated for over 40,000 people. The data cover 410 everyday activities and are provided with English action class labels.

What you can do with it

  • Training and testing human skeleton and keypoint detection models
  • Training systems to recognize everyday human actions from photographs
Tasks
Image classification, Object detection
Languages
English
Size
10K–100K rows
Downloads this month
44.8K
Likes
18

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Voxel51/MPII_Human_Pose_Dataset")