Simple (pretrained) unet training code with histopathology
Two pretrained architectures were mainly used: DenseNet-161 encoding Unet and Efficientnets
WSI slicer is in /utils
example run:
python train.py --batch_size 24 --model_type dense --n_eval 1000
check the utils/config.py for configs
data loader assumes patches for loading
validation is done using DICE(f1)-score