How to train with single modality data？

[QL111111]

Thank you for sharing this code.
I want to train on a single modality dataset. Do you have any idea how should I change this code?

[sacmehta]

We concatenate four modailities into 1, see here.

3D-ESPNet/main.py

Line 38 in edcc3b9

input = torch.cat([inp, inputA, inputB, inputC], 1) # dim-0 is batch

Instead you can feed one modality.

ALso, you need to change number of channels to 1. We use 4 for four modalities.

3D-ESPNet/main.py

Line 352 in edcc3b9

args.channels = 4 # because 4 modalities. You can think each modality as a single channel (R or G or B) of an RGB image

Obviously, you need to make changes in other files too.