a pytorch implementation of kagge pneumonia competition
Data are installed from kaggle competition 'RSNA Pneumonia Detection Challenge'
- 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
