ismail31416/LumiNet

The official (TMLR) implementation of LumiNet: Perception-Driven Knowledge Distillation via Statistical Logit Calibration

29
/ 100
Experimental

This project helps machine learning engineers create smaller, more efficient deep learning models (student models) that retain much of the accuracy of larger, more complex models (teacher models). It takes a large, pre-trained image classification model and a smaller, target model architecture as input. The output is a highly optimized, smaller model capable of performing image classification with improved accuracy compared to standard methods. It is primarily for deep learning practitioners working on computer vision tasks.

No commits in the last 6 months.

Use this if you need to deploy high-performing image classification models to resource-constrained environments by making them smaller and faster without significant accuracy loss.

Not ideal if you are not working with image classification tasks or if you do not have existing large 'teacher' models to distill knowledge from.

deep-learning computer-vision model-optimization image-classification edge-ai
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 6 / 25
Maturity 8 / 25
Community 13 / 25

How are scores calculated?

Stars

17

Forks

3

Language

Jupyter Notebook

License

Last pushed

Aug 17, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/ismail31416/LumiNet"

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