1 comment

[ 3.1 ms ] story [ 11.0 ms ] thread
Hey everyone,

for my research project, I wrote a small python code that scrapes random public screenshots from prnt.sc and uses the Moondream2 vision model (from Hugging Face) to generate descriptions of the images.

Demo here:

https://youtu.be/xEsHRepfzks

It’s fascinating (and a bit concerning) to see what kind of information can be interpreted from completely random screenshots. Let me know your thoughts!

Link to repo: https://github.com/sensahin/VisionShot