For MSDM6980
present link: https://docs.google.com/presentation/d/1Lzmad5BwLLHfSBbIZjy2d3KWjs6y3EwRXAVD71ZHRoE/edit?usp=sharing
!python train.py -l bce -e 100 -lr 1e-3 -b 4 --load '/path/to/your/model.pth'
-l/--loss: criterion for training, options are: 'bce': binary cross entropy loss, 'energy': energy loss, 'ac': active contour loss, 'nll': negative log likelihood loss, 'dice': dice loss, 'dice_bce': dice loss + binary cross entropy loss
optional: -e/--epochs: number of epochs -lr/--learning_rate: learning rate -b/--batch_size: batch size --load: load model path
!python predict.py -m '/path/to/your/model.pth' -i '/path/to/your/image' -im '/path/to/your/mask' -o '/path/to/your/pred_img'
- In energy loss: update torch version from 1.7.0- to 2.0+.
- In energy loss: