70 Dog Breed Classification

This is a Next.js app that classifies 70 different dog breeds, trained with Efficient Net B0, achieved ~90% accuracy.

The model is trained using PyTorch, can be viewed in /pytorch-training folder.

After training, the model is saved as ONNX format. The model is loaded and inferenced on edge using ONNX Runtime Web API, thus has a low edge inference time.

Tech Used

WebApp

• Next.js
• Tailwind
• ONNX Runtime Web API
• Edge Inference

Machine Learning Model

• PyTorch
• ResNet18 and EffNet B0
• Kaggle Dataset - here

Getting Started

First, run the development server:

npm run dev
# or
yarn dev
# or
pnpm dev

Open http://localhost:3000 with your browser to see the result.