caoscott/SReC
PyTorch implementation of SReC, a neural-network-based lossless image compression method that frames compression as a super-resolution problem.

Velocity · 7d
+0.5
★ / day
Trend
→steady
star history
This repository implements the SReC method for lossless image compression using neural networks. It frames lossless compression as a super-resolution problem and achieves state-of-the-art compression rates on datasets like ImageNet64 and Open Images. The project provides training, compression, and decompression utilities with pre-trained model weights released.