2025 Pet Hacks Contest: Feline Facial Recognition Foils Food Filching
1 min read
Summary
Joe Mattioni built an automatic feline feeding system that incorporated facial recognition after one of his cats, Layla, started eating the other cat’s food.
To do this, he adapted a feeding bowl with a motion-activated lid that was fitted with a camera connected to an old Android phone running a custom TensorFlow Lite model.
However, as there were no pre-existing cat facial recognition models available, he decided to use the human MobileFaceNet model as a basis and added in extra training data under the guise of 5000 feline faces.
This was done after converting the TFLite model to a Keras model using Google CoLab, and then translating it back to TFLite for android phone usage.
While this is not the easiest, fastest or most dependable solution, it is a fun learning adventure and the fact it keeps the cat on a healthy diet is a bonus.
