COVID CXRay Classification

Transfer learning for COVID, Viral, and Normal chest X-ray classification achieving 95%+ accuracy.

Overview

Implemented pre-trained state-of-the-art models (VGG16, ResNet152, InceptionV3, EfficientNetB7, MobileNet) for classifying Chest X-Ray images of COVID, Viral, and non-infected patients.

Results

  • VGG16 and MobileNet performed best with accuracy above 95%
  • Reinforced the hypothesis that pre-trained models work well for image classification tasks
  • Conducted semantic segmentation using U-Net on the dataset

Tech Stack

Python Keras TensorFlow TensorBoard OpenCV NumPy Matplotlib Pandas

Supervised by: Dr. Swakkhar Shatabda, Associate Professor, United International University