CompressAI is built on top of PyTorch and provides:

  • custom operations, layers and models for deep learning based data compression

  • a partial port of the official TensorFlow compression library

  • pre-trained end-to-end compression models for learned image compression

  • evaluation scripts to compare learned models against classical image/video compression codecs

CompressAI aims to allow more researchers to contribute to the learned image and video compression domain, by providing resources to research, implement and evaluate machine learning based compression codecs.

Model Zoo#

CompressAI includes some pre-trained models for compression tasks. See the Model Zoo section for more documentation.

The list of available models, trained at different bit-rate distortion points and with different metrics, is expected to grow in the future.