CNN Image Classifier

A deep learning project that implements a Convolutional Neural Network (CNN) for image classification using PyTorch and FastAPI. The project includes model training, evaluation, visualization, and a web interface for real-time predictions.

Features

• Model Architecture: Customizable CNN implementation with configurable layers
• Data Pipeline: Efficient data loading and preprocessing with augmentation
• Training: Configurable training loop with early stopping and checkpointing
• Evaluation: Comprehensive metrics and visualizations
• Web Interface: FastAPI-based web application for real-time predictions
• Documentation: Detailed documentation and interactive notebooks

Documentation

• Technical Documentation - Detailed technical specifications and implementation details
• API Documentation - API endpoints and usage guide
• Development Guide - Guide for developers and contributors

Project Structure

cnn-image-classifier/
├── src/
│   ├── data/           # Data loading and preprocessing
│   ├── models/         # Model architecture and training
│   ├── utils/          # Utility functions
│   └── config.py       # Configuration settings
├── app/                # FastAPI web application
├── docs/              # Documentation
├── notebooks/         # Jupyter notebooks
├── data/             # Dataset storage
├── models/           # Saved models
├── results/          # Training results and visualizations
└── logs/             # Training logs

Installation

1. Clone the repository:

git clone https://github.com/gitarber/cnn-image-classifier.git
cd cnn-image-classifier

2. Create and activate a virtual environment:

python -m venv venv
# On Windows:
.\venv\Scripts\activate
# On Unix or MacOS:
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

Usage

Training the Model

python -m src.main

This will:

• Load and preprocess the CIFAR-10 dataset
• Train the CNN model
• Generate visualizations and metrics
• Save the best model

Running the Web Interface

python -m app.main

Access the web interface at http://localhost:8000/docs

Configuration

The project is highly configurable through src/config.py:

• Dataset parameters
• Model architecture
• Training settings
• Data augmentation
• Visualization options

Results

The training process generates:

• Model checkpoints
• Training history plots
• Confusion matrix
• Feature maps
• Class activation maps

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. See our Development Guide for more details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

• Arber Shquti - GitHub