PNEUMONIA DETECTION USING CHEST X-RAYS

Abstract

Pneumonia is a serious illness that needs to be identified early and accurately in order to be treated. This project uses pre-trained models like ResNet50, VGG16, and DenseNet121 to develop a deep learning-based method for pneumonia identification. The dataset consists of X-ray pictures of the chest that have been normalised and enhanced using ImageDataGenerator. The process entails adding bespoke fully connected layers for classification and freezing the convolutional layers of the pretrained models to take advantage of their feature extraction capabilities. The dataset is used to train each model, and parameters like training and validation accuracy are monitored across several epochs. The models' performance in identifying pneumonia is highlighted by evaluating them for accuracy and loss on the test dataset.The models are saved for future use, and comparative visualizations of training and validation accuracy are presented. This study demonstrates the promise of transfer learning in medicinal image classification tasks emphasizing its applicability to pneumonia detection.