Jev Eyes Classifier takes a photo and decides which kind of scene it shows: a single person, a single child, a couple, a family with children, a person with a pet, or nobody in the picture.
Two local vision models do the looking. YOLO counts the people and pets, and SigLIP tells children from adults. Jev then makes the final choice from those facts and returns a probability for each category. A checkbox switches Jev off and falls back to plain rules, so you can compare the two.
The page text is in Turkish. It runs as a Hugging Face Space, so you open it and try it. To run your own copy you add a TypeSafe API key as a Space secret.
What you get
- A running app on Hugging Face you can use in the browser
- The Python source and the vision pipeline, readable and copyable from the Space's Files tab
Jev
Python
OpenRouter
FastAPI
React
TypeScript
CodeRabbit participant
