TransUNet
TransUNet copied to clipboard
Have you tried to train the model without pretraining on imagenet?
Comment out the load pretrain in the train.py file like this:
# net.load_from(weights=np.load(config_vit.pretrained_path)) # load pretrain