Channel mismatch error when converting the code to do multi-class segmentation. about dis

[srivastavashobhit]

The issue is resolved. Error is due to not passing the correct number of out_ch in net = ISNetGTEncoder(in_ch=4, out_ch=4).

[kabbas570]

@srivastavashobhit can you please explain how you managed to make a 4-channel label? Thanks

[srivastavashobhit]

@kabbas570
In my case, I was first incorrectly not passing out_ch=4 when creating an object of ISNetGTEncoder at line 38 of train_valid_inference_main.py

The default value of out_ch is 1. You can check this in the constructor of ISNetGTEncoder

Later, I also had to make changes to the data_loader_cache.py in im_preprocess and/or gt_preprocess to deal with 4 channels. I do not remember the changes now and do not have the files to refer.

[kabbas570]

@srivastavashobhit, thanks for the reply. Yes, I changed the gt_prrprocess file and the corresponding number of classes in ISNetGTEncoder,and fortunately, it working now,
Cheers
Abbas