A Smart Disk for In-Situ Face Recognition
Victor C. Ferreira, Alexandre Solon Nery, Felipe M. G. França · 2018
Just a couple of decades ago, significant power constraints began to shift the microprocessor design thinking process and the integrated circuit industry towards heterogeneous energy-efficient parallel systems and architectures. In this regard, Field Programmable Gate Arrays (FPGAs) play an important role as they can be used not only to prototype novel systems, but also to run various specialized parallel datapaths to more efficiently execute the critical path of an application or class of applications. However, recently the amount of devices embedded with processing capabilities and internet access has led to a substantial increase of network traffic among such different machines, prompting the development of Edge/Fog/In-Situ technologies. Thus, this work aims at designing and evaluating a smart disk system for in-situ face recognition through a Weightless artificial Neural Network (WNN) co-processor designed in High-Level Synthesis (HLS) and implemented in the system's FPGA. Although the co-processor is used for face recognition purposes, it can be used by any application that needs to use artificial neural networks. Extensive performance, circuit-area, energy consumption and network latency results shows that the co-processor can efficiently learn and recognize faces faster than the system's embedded ARM processor, while also significantly reducing network traffic.