Image Recognition System Using ANN (C#)

A simple Artificial Neural Network (ANN) implementation written in C# for classifying 5x5 binary images of digits 1 and 2.

This project was developed to understand the fundamentals of image recognition and ANN logic without using external machine learning libraries.

Overview

The system:

• Generates 5x5 matrix representations of digits 1 and 2
• Applies random pixel variations
• Converts matrices into 1D input vectors
• Trains an ANN model using a gradient-based update rule
• Evaluates classification accuracy

Key Concepts

• Artificial Neural Networks (ANN)
• Supervised learning
• Weight initialization
• Learning rate & training epochs
• Data augmentation

Technologies

• C#
• Object-Oriented Programming
• ANN implementation from scratch
• Console-based application