EfficientNet-PyTorch

EfficientNet-PyTorch copied to clipboard

Is it possible to add an input in one intermediate layer? For example:

y = self.conv_block1(image)       
y = self.conv_block2(y)        
y = self.conv_block3(y)        
y = self.conv_block4(y)
# fusion        y = y + encoded_state
y = self.conv_block5(y)       
y = self.conv_block6(y)   

You may consider loading the pretrained weights and wrapping each layer or block individually into a single nn.Module where you can play with the output of each intermediate layer/block.