Pneumonia detection model using CNN and CGAN. CNN trained for 25 epochs; CGAN for 500 epochs. Achieves 98% accuracy. Includes Gaussian filtering for noise removal, clean dataset filtering, RAG for user-driven data generation, and data augmentation for diversity and robustness. Pneumonia Detection using CNN and CGAN This project implements a deep learning-based system to detect pneumonia from chest X-ray images using a Convolutional Neural Network (CNN) and a Conditional Generative Adversarial Network (CGAN).
Features CNN Classifier trained for 25 epochs with 98% accuracy.
CGAN trained over 500 epochs to generate synthetic pneumonia-infected images and highlight disease regions.
Gaussian Filter for advanced noise tracking.
Clean Dataset Function to remove low-quality or corrupted images before training.
Region-Adaptive Generation (RAG) to generate data from user input.
Data Augmentation to improve model generalization and diversity.
Model Architecture CNN: Used for classification of pneumonia vs normal.
CGAN: Used for data generation and infected area visualization.
RAG: Enhances dataset with user-guided image region input.
Performance Accuracy: 98% on the validation set.
Supports visualization of infected areas with the CGAN module.
📁 Dataset Cleaned and preprocessed X-ray images.
Augmented using techniques like rotation, flipping, and brightness/contrast adjustment.