This repository contains an implementation of Denoising Diffusion Implicit Models (DDIM) for the superresolution problem.
- 80% of all dataset were used for training
- The remaining 20% were used for testing
- 64x64 resolution, center cropped
- 50 diffusion steps were used for testing
Input images
Output images
Target images
- Keras 3
Our code is highly inspired by clear-diffusion-keras.