unet/unet_parts.py and unet_model.py include the U-Net model.
unet/full_freq_att.py includes the proposed full-spectrum model.
utils/dataset.py is used to process dataloader.
dice_loss.py is used to compute the dice scores.
plz see the details in the following introduction.
The CAMUS echocardiography dataset is recommended !!
If you do not want to rewrite the utils/dataset.py, please using the following dataset structure.
- CAMUS |
- training >
- patinet0001 >
- img >
- img0001.png, img0002.png.....
- mask >
- img0001.png, img0002.png.....
- img >
- patinet0001 >
- training >
- testing >
As same as the training.
The name of image and its correspongding mask should be same.
run python train.py
run python predict.py