DBT_Net
DBT_Net copied to clipboard
Magic numbers in the DBTNet code
Hi, I am a little bit confused about the numbers in the following lines. If you understand it could you give a brief explanation? I appreciate your help!
co = co / 128
loss_single = torch.sum((co-gt)*(co-gt)*0.001, dim=1)
loss = loss / ((channels/512.0)*(channels/512.0))