A machine learning-powered web application built with React and TensorFlow.js that predicts dog breeds from user-uploaded images.
The project focuses on client-side ML inference, interactive UI, and a smooth, user-friendly classification workflow.
- Browser-based machine learning inference
- Image classification using pre-trained neural networks
- Interactive, real-time user feedback
- Performance-conscious client-side processing
- Clean, modern React application architecture
- React (Vite)
- JavaScript (ES6+)
- TensorFlow.js
- MobileNet (pre-trained model)
- Upload images to classify dog breeds
- Real-time inference using TensorFlow.js in the browser
- Confidence-based prediction results
- Breed search and discovery flow
- Reset and reclassification workflow
- Responsive, user-friendly interface
- Lightweight, client-side execution (no backend dependency)
- Client-side ML inference with MobileNet
- Image preprocessing and model input optimization
- React state-driven UI updates and result rendering
- Probabilistic result ranking and display
- Modular component structure for maintainability
- Optimized loading flow for model initialization and inference performance
Deployed via Netlify.
This repository represents an experimental machine learning web app, showcasing TensorFlow.js integration, client-side AI workflows, and interactive React-based UI development.
It demonstrates practical ML-in-the-browser implementation without requiring a backend server.
The app performs inference entirely in the browser — no backend required. Clone the repository and install dependencies:
npm install
npm run dev