The project is about diagnosing Pneumonia from X-ray images of lungs of a person using self laid convolutional neural network. The images were of size greater than 1000 pixels per dimension and the total dataset.

• The dataset contained 5000+ X-ray images, labelled as showing symptoms of Pnuemonia or not.
• The work includes
  • Pre-processing of data.
  • Laying down a Deep Convolutional Neural Network architecture from scratch.
• The model showed a recall of 95% and a precision of 80%.
  • In context of the problem statement, recall of the model plays a more crucial role for the successfull classifcation of images.
• The final model architecture, loss functions and regularization steps have been chosen after continous hyper parameter searches.

Automated methods to detect and classify human diseases from medical images.

• Batch Normalization
• Xavier Initialization
• Dropout Regularization
• Effective Backprop
• How Does Batch Normalization Help Optimization? (No, It Is Not About Internal Covariate Shift)