Histopathological classification of cancer via Deep Learning

Developed a web application which can in live mode detect various cancers and determine the type of cancers with an accuracy of over 99% based on the images uploaded by a user (Frontend is implemented in Python Flask, Backend is built via Keras and consists of 9 CNNs trained on clinical data). Such concepts as Ensemble Learning and Transfer Learning were applied to improve accuracy. For mobile application purposes the models were also compressed up to 2 times without loss in accuracy through channel pruning technique.