Cat Face App

I made this experimental AI-powered web app to detect cat breeds from photos. Using modern web tools and machine learning, I built it to explore client-side image classification with a custom-trained model.

Features: Upload a cat photo to identify its breed Real-time cat face detection using COCO-SSD Custom image recognition model I trained on Google Colab GPUs Fully browser-based, no server required

Built With: HTML, CSS, JavaScript (Svelte) - UI & interaction TensorFlow.js - Model loading & inference COCO-SSD - Cat face detection Google Colab - Model training with GPU acceleration

Dataset: This app was trained on the Kaggle Cat Breeds Dataset (https://www.kaggle.com/datasets/denispotapov/cat-breeds-dataset-cleared), which I cleaned and augmented by: Removing low-resolution and duplicate images. Adding more cat images sourced from Wikimedia and Flickr to enhance model diversity and robustness. Ensuring a balanced dataset to improve model accuracy.

How It Works: Upload or take a photo of a cat The app detects the cat's face using COCO-SSD A custom-trained TensorFlow model classifies the breed Results are shown instantly in the browser

Disclaimer: This is an experimental project. Accuracy may vary depending on photo quality and breed coverage in the training data.