Architecture Exploration and Customization Tool of Deep Neural Networks for Edge Devices

Abstract

Recently, Deep Neural Network(DNN)-based applications are increasing in embedded edge devices. However, due to the high computational complexity, DNN has limitations in properly executing and optimizing on edge devices. As s result, DNN exploration framework is required to customize various DNN models for edge device from the software to hardware perspectives. In this paper, we provide a GUI-based framework for architectural exploration of DNN networks in edge devices. It provides software optimization such as quantization and pruning, as well as hardware performance analysis using Virtual Platform(VP)-based Deep Learning Accelerator(DLA).