I'm wondering if there a way to enable a CNN to learn the non-linear transfer between two images. For example, the original image looks like

enter image description here

The target image looks like

enter image description here

I would like to develop a CNN that learns the pixel-wise relationship between the two images. Then, given a new image, the CNN is able to process the image to a way that is learnt.

Please suggest where I should start?

  • Well, what is the relationship between the two images? Do you know? How were they obtained? And, what have you tried? Have you tried anything so far? Generally machine learning involves trying many techniques to see which work well, so your question would be more valuable if you had already tried some approaches and could describe for us the results you've found so far. – D.W. Aug 10 at 6:07
  • The relationship between the two images are not clear. But one thing for sure is that both images represent the same objects, just their pixel values are transferred in some way that I wish machine learning algorithm could discover. – Kyle Aug 10 at 8:41
  • So, how were the images obtained? – D.W. Aug 10 at 15:59
  • The image source is unknown, and how the images were obtained are unclear. – Kyle Aug 12 at 1:03

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.