So I am currently trying to create a program that when provided a list of descriptive words or a passage of text, would create a piece of abstract art based on the feelings evoked by those words. I figured that AI would probably be the best way of achieving the results I am looking for. Even though I do have quite a bit of experience with programming, I have just about zero knowledge on the subjects of machine learning, artificial intelligence, neural nets, etc. Another issue is that I simply don't have much time to write this program and I simply don't have the luxury of thoroughly learning about this subject.

I came across a really cool python thingy called StackGAN that is specifically made for text to image generation. My plan was to take a bunch of pictures of abstract art and for each one write down a list of descriptive words or emotions that they evoke. I figure that there must be some way for me to then feed the pictures plus associated words/description into the neural net as training data but have no idea how to do that and the fact that there is very little documentation about the program doesn't help. Even after spending a few days trying to make sense of the code I am completely lost as to use StackGAN.

So.... my question is: how exactly do I set up the training data and train StackGAN to do what I am trying to do? The github's README.md mentions some sort of CNN-RNN text embedded in the images used for training, and I have no idea what this means, if its necessary, or how I would add such data because I can't find any sort of thorough instructions. Also if you have any suggestions on alternative more user friendly libraries that would be greatly appreciated.

StackGAN link: https://github.com/hanzhanggit/StackGAN-v2


Your Answer

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

Browse other questions tagged or ask your own question.