Data Augmentation using MNIST dataset

Deep learning is gaining popularity with the increase in the field of data science and the amount of data produced every single day. The process of acquiring or generating larger amount of data for the deep learning models can be expensive. To cut down the cost of producing labelled data, data augmentation is a cost-effective solution. Data augmentation is a widely accepted technique which is used to artificially increase the potential of deep learning networks. It plays a crucial role in cases where the ground truth data is limited. This method introduces various types of transformations to create new training samples for the purpose of executing machine learning models. Shorten C and his teammate presented a survey on the data augmentation techniques which included geometric transformations, erasing, cropping and many more. The goal of this project is to execute data augmentation techniques, initialise a convolutional neural network, execute it and obtain desired results.