why take first 32 channels as attention map in model.py at line 112 #5
Description
could you tell me why just take 32 channels feature map as attention map for fusioning？
{{ message }}
could you tell me why just take 32 channels feature map as attention map for fusioning？