Training & Testing

We note that the pre-trained models are provided for easier reproduction of our reported results.

CUDA_VISIBLE_DEVICES=0 python train.py --start_channel 8 --using_l2 2 --smth_labda 5.0 --lr 1e-4 --trainingset 4 --checkpoint 403 --iteration 201501

CUDA_VISIBLE_DEVICES=0 python infer_bilinear.py --start_channel 8 --using_l2 2 --smth_labda 5.0 --lr 1e-4 --trainingset 4 --checkpoint 403 --iteration 201501

python compute_dsc_jet_from_quantiResult.py

Changes

In the original arXiv draft, I used the notations LessNet_4, 6, 8, 12, and 16 (LessNet_C) to denote different variants of LessNet, assuming a starting channel of 4C that is progressively reduced to 3C, 2C, and C. However, I recently noticed that in the code implementation, the starting channels were actually set to 8×, 6×, 4×, and 2× a start channel. To align the notation with the actual implementation, the model names should be updated to LessNet_8, 12, 16, 24, and 32. But the computational results remain unchanged.

For instance, the changes in TABLE IV will be:

Was	Now	Unchanged	Unchanged	Unchanged	Unchanged	Unchanged
C	C	Parameter	Mult-Adds	Memory	Dice	J<0
4	8	44,336	152	10.31	0.749±0.040	0.808±0.389
6	12	96,264	327	15.23	0.757±0.040	0.749±0.376
8	16	168,032	570	20.16	0.761±0.039	0.742±0.353
12	24	371,088	1250	30.00	0.766±0.039	0.852±0.392
16	32	653,504	2200	39.84	0.768±0.039	0.830±0.397

Only the naming was incorrect; the values reported are correct. I have updated the code accordingly to reflect these changes.

Acknowledgment

We note that parts of the code are adopted from IC-Net, SYM-Net, and TransMorph (in chronological order of publication).

The preprocessed IXI data can be found in TransMorph.