I am trying to generate labeled training data for an artificial visual cortex to learn on. I'm looking for a tool that will, instead of rendering the scene at each pixel just give me the pre-rendered data associated with each pixel. Presumably something like:

[x,y] -> [normal, depth, color, lighting, (lens flares?), texture, etc]

  • \$\begingroup\$ Yes, it is possible. Easy? No idea. You might want to consider using an open-source game and modifying the source code. \$\endgroup\$
    – user253751
    Oct 12, 2015 at 2:13
  • 1
    \$\begingroup\$ Yes, it is possible, and someone already did it with Starcraft (pretty awesome btw) -> graphics.stanford.edu/~mdfisher/GameAIs.html \$\endgroup\$
    – glampert
    Oct 12, 2015 at 18:17
  • \$\begingroup\$ @glampert woo, thanks, hopefully this works for me. \$\endgroup\$
    – Zackkenyon
    Oct 13, 2015 at 0:47

1 Answer 1


RenderDoc can get you a lot, if not all, of that info.

I've previously used it during development to pull out my normal-maps, light maps etc. It also pulls out all the textures used in any fragment shaders etc.

  • \$\begingroup\$ This is definitely what I needed, thank you for your help. \$\endgroup\$
    – Zackkenyon
    Oct 17, 2015 at 1:27

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .