Tuberculosis (TB) Detection From Chest X-ray images

This project focuses on detecting Tuberculosis (TB) from chest X-rays using Convolutional Neural Networks (CNNs). The dataset, sourced from Kaggle, comprises 3,500 normal and 700 TB-affected X-ray images. Various preprocessing techniques, including resizing and augmentation, were applied to optimize the data for models like InceptionV3, ChexNet, and MobileNetV2. A custom CNN architecture was also developed with 2 convolutional layers and 3 dense layers. The project leveraged TensorFlow and Keras frameworks to train and evaluate the models, addressing challenges such as imbalanced data, limited dataset size, and overfitting. The primary performance metric used was accuracy.

Here is the presentation for this project. And the code can be found in this github repository.