How Can Deep Neural Networks Be Generated Efficiently for Devices with Limited Resources

摘要: Despite the increasing hardware capabilities of embedded devices, running a Deep Neural Network (DNN) in such systems remains challenge. As trend DNNs is to design more complex architectures, computation time low-resource devices increases dramatically due their low memory capabilities. Moreover, physical used store network parameters augments with its complexity, hindering feasible model be deployed target hardware. Although compressed helps reducing RAM consumption, large amount consecutive deep layers time. wide literature about DNN optimization, there lack documentation for practical and efficient deployment these networks. In this paper, we propose an generation by analyzing impact address simple comprehensive pipeline optimal deployment.