CNN Model for COVID-19 Image Classification

This project implements a Convolutional Neural Network (CNN) model to classify medical images as either COVID-19 positive or normal. The project uses Keras and TensorFlow to build and train the model, with the option to classify new images using a saved model.

Key Features

Model Architecture:

The CNN consists of multiple layers, including 2D Convolutional layers, MaxPooling, Batch Normalization, Dropout for regularization, and Dense layers for classification.

Convolution Layers:

Five convolutional layers with different filter sizes (32, 64, 96) for extracting spatial features from the images.

Pooling Layers:

MaxPooling layers are used after each convolutional layer to reduce the dimensionality and focus on important features.

Batch Normalization:

This helps in stabilizing the learning process and improves convergence speed.

Dropout Regularization:

Reduces overfitting by randomly dropping some connections during training.

Dense Layers:

Two fully connected layers at the end, with softmax activation for classification into two categories: COVID and Normal.

Model Saving and Loading:

After training, the model is saved as a JSON file (model.json) along with the weights (model.weights.h5) for future classification tasks.A separate classification function loads this saved model and predicts the label of new images.

Prerequisites

Python 3.x

TensorFlow and Keras

NumPy

Matplotlib (optional, for visualizations)