adaptivetokensampling/ATS

Adaptive Token Sampling for Efficient Vision Transformers (ECCV 2022 Oral Presentation)

41
/ 100
Emerging

This project helps machine learning engineers and researchers accelerate vision transformer models for image and video classification. It takes an existing vision transformer model and image/video data, and outputs the same model but running with significantly reduced computational cost while maintaining accuracy. It's ideal for those working with large image datasets or real-time video analysis where computational efficiency is crucial.

104 stars. No commits in the last 6 months.

Use this if you need to make your image and video classification models, specifically those using vision transformers, run twice as fast without losing accuracy.

Not ideal if you are not working with vision transformer models or if your primary concern isn't reducing computational resources like GFLOPs.

computer-vision image-classification video-analysis deep-learning-optimization machine-learning-engineering
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 16 / 25

How are scores calculated?

Stars

104

Forks

16

Language

Shell

License

Apache-2.0

Last pushed

May 03, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/adaptivetokensampling/ATS"

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