AlzAugment -THE CNN MODEL

CNN Image Classifier (4 Classes - Grayscale)

This repository contains a Convolutional Neural Network (CNN) implemented with TensorFlow/Keras to classify grayscale images into four classes (0, 1, 2, 3). It includes custom preprocessing, model training, and evaluation on unseen image data.

📁 Project Structure

The CNN uses:

• 3 Conv2D layers with L2 regularization
• Batch Normalization and MaxPooling2D after each conv layer
• Flatten and Dense layers
• Dropout (0.5) to reduce overfitting
• Final output layer with softmax for 4-class classification

##DEPENDENCIES:

• tensorflow
• keras
• numpy
• pillow
• matplotlib

Model Layers:

Conv2D(128) → BatchNorm → MaxPool
Conv2D(64)  → BatchNorm → MaxPool
Conv2D(32)  → BatchNorm → MaxPool
→ Flatten → Dense(128) → Dropout(0.5) → Dense(4, softmax)

Run the Model:

python model.py