pipixin321/HR-Pro

[AAAI 2024] Official implementation of "Point-supervised Temporal Action Localization via Hierarchical Reliability Propagation"

29
/ 100
Experimental

This project helps video analysts automatically identify and precisely locate specific actions within long, untrimmed video footage, even when only minimal starting point annotations are available. It takes raw video features and sparse 'point' labels (indicating roughly where an action starts) as input and outputs a list of identified actions with their exact start and end times. This is ideal for researchers, video content moderators, or surveillance professionals who need to efficiently analyze large volumes of video for specific events.

No commits in the last 6 months.

Use this if you need to pinpoint actions in videos but have very limited time or resources to create detailed, frame-by-frame annotations for training.

Not ideal if you already have fully labeled videos with precise action boundaries for training, as this method is optimized for sparse supervision.

video-analysis action-recognition content-moderation event-detection surveillance
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 8 / 25
Maturity 16 / 25
Community 5 / 25

How are scores calculated?

Stars

42

Forks

2

Language

Python

License

MIT

Last pushed

Mar 18, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/pipixin321/HR-Pro"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.