Classifying TB

An application that uses CNN, to predict TB.

We used a pretrained CNN(VGG16) for predictions and applied transfer learning to re-train some of it's layers on Lung X-ray(s).Apart from the VGG16, we have also added 3 dense layers for the CNN to properly learn the features. A dropout layer is also added to prevent the Neural Network from over-fitting.

Specifications


  • Model Name: VGG16
  • Optimizer: Adam
  • Number of epochs trained on: 100
  • Learning rate: 0.0001
  • Validation loss: 0.655
  • Validation accuracy: 0.78
  • Sensitivity: 0.85
  • Specificity: 0.76
  • More Information

Upload X-ray



default.png