won548 / LDCT-Octave

Low-Dose CT Denoising using Octave Convolution

LDCT-Octave

Low-Dose CT Denoising using Octave Convolution

Training

Train with U-Net

CUDA_VISIBLE_DEVICES=0 python train.py --model=unet --test_fold=10 --scheduler

Train with REDCNN

CUDA_VISIBLE_DEVICES=0 python train_ct.py --model=redcnn --test_fold=10 --scheduler

Train with Octave convolution

CUDA_VISIBLE_DEVICES=0 python train_ct.py --model=cnn_oct --test_fold=10 --scheduler --alpha=0.75

Test

For test, you need to specify some arguments.

--test_date: the directory where your experiment located.
--test_epoch: the epoch of your test
--desc     : description of your experiment. your predictions (image, nitfi, plots will be saved the folder given description)

Test with redcnn

python test_ct.py --model=redcnn --test_date=$TEST_DATE --test_epoch=best --subject=L506 --desc="redcnn"

Test with Octave convolution

python test_ct.py --model=cnn_oct --alpha=$ALPHA --test_date=$TEST_DATE --test_epoch=best --subject=L506 --desc="cnn_oct"