H&E slides color normalisation using generative adversarial networks
Academic project
Code based on StainGAN rep: https://github.com/xtarx/StainGAN
I'm using the public MITOS-ATYPIE141 dataset (see more at: https://mitos-atypia-14.grand-challenge.org/).
The original dataset consists of 284 frames at x20 magnification selected in breast cancer biopsy slides by the team of Professor Frédérique Capron, head of the Pathology Department at Pitié-Salpêtrière Hospital in Paris, France. Each slide was stained with standard hematoxylin and eosin (H&E) dyes and was scanned by 2 scanners: Aperio Scanscope XT and Hamamatsu Nanozoomer 2.0-HT.
To run training and testing, just download the train_inference.ipynb notebook and open it with Google Colab.
In the notebook, we first download the data, then run training and/or inference using the respective config files.
