Is there a particular section of image data which when trained on deepdream algorithm and given some input image produce a resulting image from which we can conclude that the deepdream can be used for this image set (other than just producing hallucinations on input image for fun) ?

  • 4
    $\begingroup$ I can't actually understand what you're asking. I think it would be helpful if you broke your question down into more than one sentence. $\endgroup$ – David Richerby Jun 4 '16 at 21:18
  • $\begingroup$ According to my understanding deepdream algorithm is trained on a set of images (say suppose dogs) and when we give the input image of sky, it detects the features of image that are similar to dogs and amplifies it. I wanted to know if there is any image set that when we train on deepdream produces results that can be used for any other application. Or is there any other useful application of deepdream ? I am not sure if my understanding is right, so kindly corrct me if i'm wrong $\endgroup$ – kavya jalla Jun 4 '16 at 22:52
  • 3
    $\begingroup$ As @DavidRicherby suggested, it would be helpful to clear things a bit, but by editing the question, not the comment. Also I am afraid that your question can be downgraded to "I have the image filter X, can I do anything reasonable (or just different then intended art effect) with it?", which is not clear objective. $\endgroup$ – Evil Jun 5 '16 at 0:01
  • $\begingroup$ By not clear objective I mean that there might be a dozen of applications - and there is no way of telling which one is better, a.k.a opinion based... $\endgroup$ – Evil Jun 5 '16 at 0:05
  • 1
    $\begingroup$ You ask three different questions in the title alone, and I can only second David's concerns. Please tighten up your post to ask one clear question. (You may have to do some research in order to do so.) $\endgroup$ – Raphael Jun 5 '16 at 14:09

Google deepdream can be extended and has been extended in many ways as below:

  1. Image segmentation (link)
  2. Beautifying pictures/photos and human faces and bodies
  3. Image upscaling
  4. Greyscale to Color
  5. Censoring and go on..

Here, somebody using MIT's places data to train the deepdream (video)

Most of the problems in the list are at amateur stage still its one step to perfection.

If interested, you should play/improve/contribute with the code here GitHub


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