DEPLOYMENT.md

Intel Image Classifier - Deployment Guide

Overview

This application is a full-stack web classifier for natural scene images using two CNN models:

• PyTorch Model: Custom CNN architecture
• TensorFlow Model: Custom CNN architecture

The application combines:

• Backend: Django REST API
• Frontend: React with Material-UI
• Models: Two Deep Learning models for image classification

Quick Start for Hugging Face Spaces

Prerequisites

• Git
• Docker & Docker Compose (for local development)
• Or access to Hugging Face Spaces

Option 1: Deploy to Hugging Face Spaces (Recommended)

1. Fork/Clone the Repository

   git clone https://github.com/danielle2035/Intel_classification.git
   cd Intel_classification
   

2. Add Your Trained Models

   Place your trained model files in the backend/api/models/ directory:

   backend/api/models/
   ├── pytorch_model.pth      (PyTorch model)
   └── model_best.keras       (TensorFlow model)
   

3. Push to Hugging Face

   # Add HF as remote
   git remote add hf https://huggingface.co/spaces/YOUR_USERNAME/Intel_classification
   
   # Push to deploy
   git push hf main
   

4. Access Your App

   • Go to: https://huggingface.co/spaces/YOUR_USERNAME/Intel_classification
   • The app will build and deploy automatically!

Option 2: Build and Run Locally

With Docker Compose (Separate Services)

docker-compose up --build

Services will be available at:

• Frontend: http://localhost:3000
• Backend API: http://localhost:8000
• API Docs: http://localhost:8000/swagger

With Docker (Unified Container - HF Mode)

docker build -t intel-classifier .
docker run -p 7860:7860 intel-classifier

Access at: http://localhost:7860

Without Docker (Development)

1. Backend Setup

   cd backend/api
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r ../requirements.txt
   python manage.py migrate
   python manage.py runserver 0.0.0.0:8000
   

2. Frontend Setup (separate terminal)

   cd frontend
   npm install
   npm start
   

3. Access

   • Frontend: http://localhost:3000
   • Backend API: http://localhost:8000

API Endpoints

Classification

POST /api/classify/

Classify an image using either PyTorch or TensorFlow model.

Response:

{
  "class": "mountain",
  "confidence": 0.95,
  "model_used": "pytorch",
  "probabilities": {
    "buildings": 0.02,
    "forest": 0.01,
    "glacier": 0.01,
    "mountain": 0.95,
    "sea": 0.01,
    "street": 0.00
  }
}

Models & Classes

Supported Classes

• buildings / Bâtiments / Kër yi
• forest / Forêt / Géej bu wees
• glacier / Glacier / Dëkk bu sedd
• mountain / Montagne / Tund bi
• sea / Mer / Géej bi
• street / Rue / Yoon bi

Deployment Checklist

• Add trained models to backend/api/models/
• Update ALLOWED_HOSTS in settings if needed
• Test locally with Docker
• Push to Hugging Face Spaces
• Test on HF Space URL

Troubleshooting

Port Already in Use

lsof -i :7860  # Find process
kill -9 <PID>   # Kill it

Models Not Loading

• Ensure files are in backend/api/models/
• Check file names: pytorch_model.pth, model_best.keras

CORS Errors

• Verify backend and frontend are accessible
• Check Django CSRF_TRUSTED_ORIGINS includes your HF URL

Performance Tips

1. Resize images before upload (< 10MB)
2. PyTorch is generally faster on CPU
3. Adjust confidence threshold in api_views.py if needed