Hi, I am working on an image reconstruction problem, where images in my dataset are of size (1, 1, 128,128 ) To improve the current results I came across Squeeze and Excitation block, which allows the network to feature recalibration (emphasize informative features and suppress less useful features)
however, Squeeze and Excitation network works with an image channel greater than 1. is there any possibility that I can still use this network by converting 1 channel to a 3 channel image.
Has anyone worked with SE network here? any advice would be a great help.
