Could you pick me out as a face in a crowd? Could you do it as well as AI? This post explores my learnings in a project where I used DeepFace to partition about 4000 photos in photos of me / not me.
In this post I use a combination of OpenAI's CLIP model, PyTesseract, and Principal Component Analysis to identify, crop, and downsample pictures of text labels to the most pertinent content.
