CppCatDogClassifier

An end-to-end C++ implementation of a neural network built from scratch for detecting and classifying images of dogs and cats.

Warning

This software is unfinished. Keep your expectations low.

Table of Contents

• CppCatDogClassifier
  • Table of Contents
  • Introduction
  • Project Structure
  • Getting Started
    • Prerequisites
    • Compiling and Running
  • Implementation Details
  • Future Enhancements
  • License
  • Acknowledgements

Introduction

This project aims to provide a basic understanding of neural networks by creating a feed-forward neural network in C++ without relying on third-party libraries. The neural network is trained using a dataset of 28x28 grayscale images to detect if an image contains a dog or a cat.

Project Structure

NeuralNetworkProject:
|-- data:
|   |-- images:
|   |   |-- cat1.bmp
|   |   |-- cat2.bmp
|   |   ...
|   |-- labels:
|   |   |-- cat1.txt
|   |   |-- cat2.txt
|   |   ...
|-- src:
|   |-- ActivationFunction.cpp
|   |-- LoadImages.cpp
|   |-- main.cpp
|   |-- Matrix.cpp
|   |-- NeuralNetwork.cpp
|-- include:
|   |-- ActivationFunction.hpp
|   |-- Matrix.hpp
|   |-- NeuralNetwork.hpp

Getting Started

Prerequisites

1. Ensure you have a C++ compiler installed, preferably GCC.
2. Prepare your dataset:
   • Place BMP images (28x28 pixels and 24-bit format) in the data/images directory.
   • Place corresponding labels in the data/labels directory.

Compiling and Running

1. Navigate to the NeuralNetworkProject directory.
2. Compile the project using the following command: ```bash g++ -o NeuralNetwork -Iinclude src/*.cpp -std=c++11 ```
3. Execute the compiled program: ```bash ./NeuralNetwork ```

Implementation Details

• Matrix: Handles matrix operations vital for the neural network.
• ActivationFunction: Defines the sigmoid activation function and its derivative for neuron activations.
• NeuralNetwork: Establishes the neural network's structure and operations, including feed-forward and backpropagation.
• LoadImages: Contains functions to load BMP images and their corresponding labels.

Future Enhancements

• Improve neural network architecture with convolutional layers.
• Implement mini-batch gradient descent.
• Split the dataset into training, validation, and test sets.
• Utilize a build system like make or CMake for efficient compilation.

License

This project is open-source and available under the MIT License.

Acknowledgements

• The Kaggle Dogs vs. Cats dataset for initial training data inspiration.