Cat Breed Classification Web App Utilizing Teachable Machine

Abstract

Cats are widely kept as pets and exhibit diverse physical traits, making breed identification difficult, especially when breeds look similar or image quality is limited. This research develops a web-based cat breed recognition system built with Google’s Teachable Machine and trained on images from 15 different breeds. ReactJS is used as the interface layer, while Flask handles the backend operations and model integration. The workflow includes training the model with internal images and validating its performance using external samples to assess robustness. Experimental results indicate that the system can recognize cat breeds with strong performance, reaching 99.52% accuracy for Persian cats and maintaining reliable predictions even when tested on low-quality or imperfect images. Overall, the system offers a practical solution for enthusiasts and researchers to more easily and accurately distinguish various cat breeds.