How could I joint feature maps of different dimension

with different kernel sizes ,feature maps also have different size ,like (1,1,62,62)and (1,1,67,67), in such condition, torch.cat doesn’t work as their dimensions don’t match. is there any way can make them together? Thanks a lot

I don’t think so :confused: You can always flatten them to (1,-1) where -1=62x62 or 67x67.

Otherwise you’re stuck with padding or cropping the feature maps

it works. great appreciated

1 Like