DatingClaw
Visual research, 2026. By Rivian Pratama. How to make a machine understand human visual taste.
Visual taste is one of the most personal things a person has and it is also one of the hardest to explain. This project asks whether a machine can learn that taste from the choices we make.
The pipeline extracts face embeddings with InsightFace and image embeddings with CLIP. It scores each profile with a KNN over labeled references, supervised regressors, and a preference classifier that predicts swipe intent. The system grew from a face-similarity test into almost 10,000 labeled examples, and the key finding is that benchmark accuracy does not equal live satisfaction.
I wanted to know if a machine could ever learn what my eyes already know.