Custom CNN Model for Pheumonia Detection using Chest XRays

This is a personal project

Project Overview:

The project focuses on developing a Neural Network from scratch without any helper libraties. Just Maths and logic.

Key Highlights:

• Implemented a Single Hidden Layer Neural Network for MNIST Dataset
• Wrote the functions for Forward Pass, Backpropagation, Weights Updation, and all the helper functions.
• Achieved accuracy of ~86% (Proof of Maths).
• Shifted from a fixed single layer architecture to a modular, robust architecture.
• Used He initialization for initialization of weights (Best suited method for while using ReLU activation.)

This work not only strengthened my foundations in Deep Learning and Neural Networks, but also helped me better understand the mathematics behind all of it.

Hand Written Notes explaining the Maths: