Light-Weight RetinaNet for Object Detection on Edge Devices

Yixing Li, Akshay Dua, Fengbo Ren

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper aims at reducing computation for Retinanet, an mAP-30-tier network, to facilitate its practical deployment on edge devices for providing IoT-based object detection services. We first validate RetinaNet has the best FLOP-mAP trade-off among all mAP-30-tier network. Then, we propose a light-weight RetinaNet structure with effective computation- accuracy trade-off by only reducing FLOPs in computationally intensive layers. Compared with the most common way of trading off computation with accuracy-input image scaling, the proposed solution shows a consistently better FLOPs-mAP trade-off curve. Light-weight RetinaNet achieves a 0.3% mAP improvement at 1.8x FLOPs reduction point over the original RetinaNet, and gains 1.8x more energy-efficiency on an Intel Arria 10 FPGA accelerator in the context of edge computing. The proposed method potentially can help a wide range of the object detection applications to move closer to a preferred corner for a better runtime and accuracy, while enjoys more energy-efficient inference at the edge.

Original languageEnglish (US)
Title of host publicationIEEE World Forum on Internet of Things, WF-IoT 2020 - Symposium Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728155036
DOIs
StatePublished - Jun 2020
Event6th IEEE World Forum on Internet of Things, WF-IoT 2020 - New Orleans, United States
Duration: Jun 2 2020Jun 16 2020

Publication series

NameIEEE World Forum on Internet of Things, WF-IoT 2020 - Symposium Proceedings

Conference

Conference6th IEEE World Forum on Internet of Things, WF-IoT 2020
Country/TerritoryUnited States
CityNew Orleans
Period6/2/206/16/20

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Networks and Communications
  • Hardware and Architecture
  • Information Systems and Management
  • Statistics, Probability and Uncertainty
  • Computational Mechanics
  • Instrumentation

Fingerprint

Dive into the research topics of 'Light-Weight RetinaNet for Object Detection on Edge Devices'. Together they form a unique fingerprint.

Cite this