Tutorial: Deploying Deep Learning Networks

The interesting aspect of this chapter of the book is in deploying deep learning networks. This section provides hands-on experience for designing and deploying deep learning networks. This tutorial is designed to handle workflow from data set creation, the deep learning model design, training the deep learning model, testing the deep learning model, and deploying the deep learning model in Internet of Things (IoT) edges and also in cloud native infrastructure. Moreover, the focus is on bringing out issues involved in deploying trained deep learning networks in IoT edges for real-time inference. During inference, sub systems are used to collect real-time input from respective sources and perform inference on given input data. Moreover, the mentioned items are expected to work in embedded hardware with limited resources. Loading a trained deep learning network onto embedded systems is a challenging task and in this regard, many companies appear to be providing their own solution, which might fit into their version of silicon devices, but may not be good for those of other companies for performing inference. The tutorial uses documents from Google Drive so that a learner can refer to a resource document online and make quick progress in learning how to deploy “deep learning networks in an edge.” The URL of a given resource is associated with a QR code. The following is the QR code for URLs that are used in the tutorial. The training deep learning model is in item 2 and it requires CPU+GPU hardware such that accelerated computing is feasible. Also, the mentioned training tasks can be cloud native as well or it can be done on-frame. Obtaining a data set is handled in item 1. Handling item 1 requires domain knowledge on a given data set creation.