Conditional Diffusion Model: A Pytorch Implementation Denoising Diffusion Probabilistic Models (DDPMs) are generative models that utilize forward and reverse noising processes. Recently, diffusion models have been applied to image compression tasks by incorporating conditionals into a denoising decoder.

In this notebook, I implement a fundamental principle of conditional diffusion models as presented in Lossy Image Compression with Conditional Diffusion Models (https://proceedings.neurips.cc/paper_files/paper/2023/file/ccf6d8b4a1fe9d9c8192f00c713872ea-Paper-Conference.pdf). The Fashion-MNIST dataset is used in this tutorial.