Input pipeline

Libraries to implement deep learning architecture has made it easy for researchers or developers to implement their own architecture and evaluate the performance on tasks at hand. Majority of the time goes in designing a pipeline which takes the data in a raw format and structured the data to be able to feed it for model training. Script input.py details the input pipeline using tensorflow queuing mechanism for training neural networks/deep neural networks. Advantage to this approach is reduced time for fetching mini-batches in each iteration, ability to shuffle data in each epoch (bringing more diverse samples for training).