testing performance

[swestland]

This is really helpful code and very good documentation on how to use it.

I just have one small question. I trained the network on some images (I used just 8 images as a test). Based on the visual performance during training the network was producing great outputs after 100 epochs. However when I then tested the trained network on the same images I used in training - exampleOutput = p2p.translate(p2pModel, exampleInput) - the output was not as good as I expected. The quality of the output images was similar to what the network had been producing after about 60 epochs during training.

Is there something I am missing here that resulted in this behaviour?
Steve

[justinpinkney]

Hmm, the only thing I can think which might be the difference is that there is an "ARange" options to p2p.translate which defaults to 255, but this might not be correct for your input data.