CompressAI
CompressAI copied to clipboard
Implementation of GMM for Cheng2020
I made a Gaussian Mixture Model implementation for the two Cheng2020s. If anyone else is interested can make a pull request after I clean up the code a bit.
Not sure about the vanilla models, but when I experimented with parallel context models (Checkerboard, etc) GSM or GMM actually made a big difference in model performance. So I hope someone will find this interesting.
PRed to #239 in case anyone is interested