I have set of night images which I will be using for self driving. But I want to convert those images into day images. I have developed algorithm based on day image but it is not good for night images , so I want to convert night images to day images then feed into the network. As far as I have explored image colourization techniques of grey scale image ( converting night image to black and white and then coloring it) ànd night to day style transfer using GANs. Is it possible to achieve the goal with the above two mentioned techniques or you can also suggest any other way.

Example link. :- https://goo.gl/images/kTwnHB


One problem with this is that dark images simply contain less information. Anyone with a background in photography will tell you it’s easier to decrease exposure on a bright image than increase exposure on a dark one (you can’t create what’s not there but you can throw away information you have). If you want all images on the same playing field, perhaps do the reverse. Or, try to extract only edges or high contrast regions from all images. Or, model them differently.

|improve this answer|||||
  • $\begingroup$ Yes I agree....I have figured it out...means I have read a research paper( learning to see in dark) in which they were able to get bright images from almost dark image but I am messed with up dataset...how can I create my dataset...the model requires one to one correspondence as it is supervised learning....but I have day image....so how can I convert to night images ? Is there any simulator that can help me...or also you can suggest any other way $\endgroup$ – Berry May 11 '18 at 6:46
  • $\begingroup$ Can't we use the day image and decrease it's brightness and use it as night image, i know its just blindly hitting the stone...can anyone find any flaw into this method for dataset conversion ? $\endgroup$ – Berry May 11 '18 at 9:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.