I Built a Pet Emotion Classifier With Python in a Weekend (Here's What I Learned)

My dog Biscuit has three moods: hungry, suspicious, and vibing. At least, that's what I thought until I started logging his behavior and running it through a classifier. What I found was... humbling. This is the story of how I built a lightweight pet emotion detection pipeline over a weekend, what worked, what didn't, and how it accidentally turned into a real product idea. The Problem (Yes, There Is One) Pet owners worry. A lot. According to the American Pet Products Association, 67% of U.S. households own a pet — and a huge chunk of them regularly Google things like "why is my cat staring at the wall" or "dog suddenly scared of nothing". The real problem isn't weird behavior. It's anxiety without context. Owners see a symptom, spiral, and either over-medicate or under-react. What if a simple classifier could give them a starting point? The Stack Python 3.11 FastAPI for the inference endpoint TensorFlow Lite (MobileNetV2 fine-tuned on pet behavior frames) OpenCV for frame extraction F