MNIST Classification with Custom Dense Layer (TensorFlow)

This project demonstrates building a neural network from scratch using TensorFlow and a custom Dense layer (SimpleDense). The model classifies handwritten digits from the MNIST dataset and achieves ~98% accuracy.

---

Project Overview

• Dataset: MNIST handwritten digits
• Goal: Classify digits (0–9) from 28x28 grayscale images.
• Key Feature: Custom implementation of a Dense (fully connected) layer.

---

Results

• Training accuracy: ~98% after 5 epochs

Requirements

• Python 3.8+
• TensorFlow 2.x

Install dependencies:

pip install tensorflow

How to run

1. Clone the repo

git clone https://github.com/Aatman09/MINIST-Custom-Layer.git
cd mnist-custom-dense

2. Run the Script

python custom_layer_correct.py