pneumonia-detection-pytorch

a pytorch implementation of kagge pneumonia competition

DataSet

Data are installed from kaggle competition 'RSNA Pneumonia Detection Challenge'

Main Idea

• 3-class classification
  • lung opacity
  • normal
  • not normal / no lung opacity
• trian with densenet121
  • fine-tuning classifier
• Grad_CAM unsupervised learning
  • improve explainability
  • thresholding heatmap to output bbox