humansensinglab/POET-continual-action-recognition

Code for the paper: "POET: Prompt Offset Tuning for Continual Human Action Adaptation" (ECCV 2024, Oral)

25
/ 100
Experimental

This tool helps you quickly teach an AI system to recognize new human actions or gestures, even with very few examples. It takes existing AI models trained on a set of known actions and allows you to add new actions without retraining the entire system. For example, if your system can recognize walking and running, you can easily add 'waving' or 'pointing' by providing just a few samples of the new actions. This is ideal for researchers or engineers working on human-computer interaction, surveillance, or fitness tracking applications who need to adapt AI models on the fly.

No commits in the last 6 months.

Use this if you need to continually expand the set of actions or gestures your AI system can recognize, using minimal new data for each addition and without forgetting previously learned actions.

Not ideal if you are looking for a system to train an action recognition model from scratch, as this tool focuses on efficiently adapting existing models.

human-action-recognition gesture-recognition continual-learning human-computer-interaction activity-monitoring
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 5 / 25
Maturity 8 / 25
Community 10 / 25

How are scores calculated?

Stars

14

Forks

2

Language

Python

License

Last pushed

Apr 25, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/humansensinglab/POET-continual-action-recognition"

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