Question; I'm trying to implement simplenet in tensorflow and I have a question that I can't seem to answer myself. The implementation I'm basing this off of is here: https://github.com/Coderx7/SimpleNet_Pytorch/blob/master/models/simplenet.py
Now here's the thing; cifar10 data is 32 by 32. There are 5 max pooling layers which each half the feature map output size. That means that by the final convolutional layer the feature map is 1 by 1. Yet the final convolutional layer uses 3 by 3 kernels for some reason!
This doesn't make a whole lot of sense to me. Can anyone explain this? I'm really stumped. If the feature map is literally 1 by 1 what possible use could the network have for 3 by 3 kernels?
EDIT: Still haven't been able to figure it out. I really am stumped. If anyone has any input it'd be appreciated.
EDIT2: I still haven't figured it out. As far as I can tell this isn't the same implementation as the original one in cafee.