The idea is to develop a web application that predicts breast cancer types (malignant or benign) upon uploading tumor images.Similarly used a traditional approach developing my own CNN architecture and got 80% accuracy.In my second approach is used transfer learning(ResNet and MobileNet) which is a pretrained model with frozen weights and used data augmentation(flip,crop and adjust the color of the image) and showered with 90% accuracy.Third approach is batch normalization then changed the optimizer ( Adam) and added regularization and dropout.Then converted it into .h5 model and deployed in a web application.