This is an unofficial implementation of the paper "DiffuseIR: Diffusion Models for Isotropic Reconstruction of 3D Microscopic Images"
Ubuntu 22.04 LTS, Python 3.9, PyTorch 2.4.0, CUDA 11.8
Since DiffuseIR shares a similar idea with Repaint, which only modified the reverse process in inference for conditional guidance, we basically include the dataloader and test code for electron microscope images.
For training, you may refer to the original Improved Diffusion repository, once you have a trained unconditional diffusion model, you can use the test code my_test.py to generate images with conditional guidance.