A large codebase would require many pictures. Maybe it's possible if you took a video and figured out a way to optimally scroll the data so as to reconstitute the codebase via AI / OCR, but that's a lot more work than most people are willing to put into it.

The goal isn't to stop all possible avenues of IP extraction, just to make the barrier a bit higher than git clone / git push.