NVTC is a vector quantization based lossy coding framework. Our paper is published at CVPR 2023.

[arXiv] [CVF Open Access]

In this repository, you will find:

• Coding playground of various toy source coding models, [here].
• Training/testing code of our NVTC models for lossy image coding, [here].

If you find this repo useful, please cite:

@inproceedings{feng2023nvtc,
  author={Feng, Runsen and Guo, Zongyu and Li, Weiping and Chen, Zhibo},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, 
  title={NVTC: Nonlinear Vector Transform Coding}, 
  year={2023},
  pages={6101-6110},
  doi={10.1109/CVPR52729.2023.00591}
}

• Tensorflow Compression: a library contains data compression tools for TensorFlow.
• CompressAI: a PyTorch library and evaluation platform for end-to-end compression research.
• PyTorch Lightning: a deep learning framework to pretrain, finetune and deploy AI models.