Skip to content

Dataset · MCG-NJU

Spatio-temporal sports action annotation dataset

MCG-NJU/SportsAction

MultiSports video dataset featuring multi-person action annotations in space and time.

What's inside

Based on sports event videos and English text annotations marked for spatial and time action localization. The dataset contains between 1,000 and 10,000 examples covering multi-participant scenarios in the frame. Created as a benchmark to overcome the limitations of simpler atomic action datasets.

What you can do with it

  • Training video spatio-temporal action detection models
  • Evaluating video analytics and athlete tracking systems
  • Training object detectors and dynamic scene classifiers
Tasks
Image classification, Object detection, Other
Languages
English
Size
1K–10K rows
Downloads this month
3.6K
Likes
51
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("MCG-NJU/SportsAction")