Integrating machine learning techniques with edge computing devices powered by Graphics Processing Units and Tensor Processing Units has revolutionized computer vision and real-time tracking systems. Object detection and motion tracking, crucial for applications like autonomous vehicles and surveillance, have advanced significantly due to this synergy. These devices provide essential computational power, thanks to their parallel architecture and memory bandwidth. Coupled with further optimization techniques such as compression, quantization, and pruning, they enable high-performance computing at the edge. However, energy consumption has become a concern due to the high computational demands of machine learning models. In this paper, we investigated energy consumption in high-performance edge devices for object detection and motion tracking, using Google Coral AI and Nvidia Jetson Nano as representative hardware platforms. In particular, we analyzed factors influencing energy expenditure, including hardware specifications, tracking algorithms, and parameter configurations. In particular, we found that the power consumption of tracking algorithms can affect the total energy budget from 9% up to almost 300% depending on the algorithm and on its configuration.

A Study on the energy-efficiency of the Object Tracking Algorithms in Edge Devices

Giacomo Di Fabrizio;Lorenzo Calisti;Chiara Contoli;Nicholas Kania;Emanuele Lattanzi
2024

Abstract

Integrating machine learning techniques with edge computing devices powered by Graphics Processing Units and Tensor Processing Units has revolutionized computer vision and real-time tracking systems. Object detection and motion tracking, crucial for applications like autonomous vehicles and surveillance, have advanced significantly due to this synergy. These devices provide essential computational power, thanks to their parallel architecture and memory bandwidth. Coupled with further optimization techniques such as compression, quantization, and pruning, they enable high-performance computing at the edge. However, energy consumption has become a concern due to the high computational demands of machine learning models. In this paper, we investigated energy consumption in high-performance edge devices for object detection and motion tracking, using Google Coral AI and Nvidia Jetson Nano as representative hardware platforms. In particular, we analyzed factors influencing energy expenditure, including hardware specifications, tracking algorithms, and parameter configurations. In particular, we found that the power consumption of tracking algorithms can affect the total energy budget from 9% up to almost 300% depending on the algorithm and on its configuration.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11576/2724531
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact