Problem
Traditional pet identification methods—such as microchips, tags, or collars—are fundamentally flawed. Tags get lost, collars fall off, and microchips require specialized scanning hardware that average citizens do not possess.
Why it matters
When a dog goes missing, the friction to identify them is too high, leading to overcrowded shelters and permanently lost pets. A frictionless identification method is needed.
Insight
A dog's nose print is as mathematically unique as a human fingerprint. By leveraging computer vision on standard smartphone photographs, we can extract distinct biometric markers, creating a non-invasive, universally accessible identification system.
Solution
Designed a microservices architecture separating heavy CV processing from the frontend. Developed a Python/FastAPI backend using OpenCV for feature extraction, and a React web app for the user workflows.
Technology used
Python, FastAPI for the backend API, React for the client interface, and OpenCV for image preprocessing and biometric vector extraction.
Impact / Result
Successfully launched the MVP phase, establishing the foundational architecture. Currently in the data collection and calibration phase to fine-tune matching accuracy across different breeds and lighting conditions.