Image Classification using CNN on CIFAR-10

A deep learning project that implements a Convolutional Neural Network (CNN) for image classification using the CIFAR-10 dataset. This project demonstrates the implementation of a robust image classification pipeline using TensorFlow and Keras.

Project Structure

.
├── data/                   # Dataset and data processing scripts
│   ├── README.md           # Dataset documentation
│   └── download.py         # Script to download the CIFAR-10 dataset
├── src/                    # Source code
│   ├── cnn_model.ipynb     # Main Jupyter notebook with model implementation
│   ├── models/             # Directory to store trained models
│   └── *.jpg               # Example images for testing
├── tests/                  # Test files
│   └── test_download.py    # Tests for data download functionality
├── Pipfile                 # Python dependencies
└── Pipfile.lock            # Locked dependencies

Features

• Data Augmentation: Implements various augmentation techniques including random flips, rotations, translations, and zoom
• CNN Architecture: Custom CNN model with multiple convolutional and dense layers
• Training Pipeline: Complete training loop with learning rate scheduling
• Model Evaluation: Includes accuracy, loss visualization, and confusion matrix
• Model Persistence: Save and load trained models for inference

Live Demo

Link = https://imageclassificationsplintersword.streamlit.app/

Requirements

• Python 3.8+
• TensorFlow 2.x
• NumPy
• Matplotlib
• scikit-learn
• Jupyter Notebook (for development)

Setup

1. Clone the repository:

   git clone https://github.com/SplinterSword/Image_Classification.git
   cd Image_Classification

2. Install dependencies using Pipenv:

   pip install pipenv
   pipenv install
   pipenv shell

3. Download the CIFAR-10 dataset:

   python data/download.py

Usage

1. Open and run the Jupyter notebook:

   jupyter notebook src/cnn_model.ipynb

2. The notebook is organized into the following sections:

   • Environment setup and imports
   • Data loading and preprocessing
   • Model architecture definition
   • Model training with data augmentation
   • Model evaluation and visualization
   • Saving and loading the trained model

Model Architecture

The CNN model consists of:

• Multiple convolutional blocks with batch normalization
• Max-pooling layers for downsampling
• Dropout layers for regularization
• Dense layers for classification
• Learning rate scheduling for better convergence

Results

The model achieves competitive accuracy on the CIFAR-10 test set. Detailed performance metrics and visualizations are available in the notebook.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.