TensorFlow Implementation of Multi-Realism Image Compression (Agustsson et al. CVPR 2023)

[Nikolai10]

Dear TFC-Community,

I have recently been working on a TF implementation of Multi-Realism Image Compression (Agustsson et al. CVPR 2023):
https://github.com/Nikolai10/MRIC.

Multi-Realism Image Compression (MRIC) can be considered the successor of the widely-known and long-standing previous state-of-the-art generative compression method HiFiC. MRIC not only significantly pushes the frontier of achievable distortion-realism but also ships with a novel beta-conditioning mechanism that allows users to choose at the decoder side whether they prefer reconstructions with high perceptual quality (generative compression), low-distortion (traditional compression) or anything in between. The beta-conditioning mechanism is an important step towards increasing the general acceptance of generative compression methods, which have often been criticized for their lack of transparency of the underlying image synthesis process.

In my implementation, I have taken some care to closely follow the naming conventions presented by Minnen et al. without changing the overall structure. I have also included a Google Colab demo, which you can find here.

I hope this work will be useful for further exploration and learning.

Kind regards,
Nikolai