EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Simplification of Deep Neural Network-Based Object Detector for Real-Time Edge Computing

Kyoungtaek Choi iD, Seong Min Wi, Ho Gi Jung iD, Jae Kyu Suhr iD

Sensors · 06.04.2023 · 23 (7) · 3777

DOI: 10.3390/s23073777 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 06.04.2023. Verfasst von Kyoungtaek Choi und 3 weiteren Autor:innen. Beteiligte Einrichtung: Department of AI Automation Robot, Daegu Catholic University, 13-13 Hayang-ro, Hayang-eup, Gyeongsan-si 38430, Gyeongsangbuk-do, Republic of Korea. Die Arbeit wurde bisher 20-mal zitiert (Crossref) und nennt 70 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Research Foundation of Korea, Ministry of Science ICT and Future Planning.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a method for simplifying and quantizing a deep neural network (DNN)-based object detector to embed it into a real-time edge device. For network simplification, this paper compares five methods for applying channel pruning to a residual block because special care must be taken regarding the number of channels when summing two feature maps. Based on the comparison in terms of detection performance, parameter number, computational complexity, and processing time, this paper discovers the most satisfying method on the edge device. For network quantization, this paper compares post-training quantization (PTQ) and quantization-aware training (QAT) using two datasets with different detection difficulties. This comparison shows that both approaches are recommended in the case of the easy-to-detect dataset, but QAT is preferable in the case of the difficult-to-detect dataset. Through experiments, this paper shows that the proposed method can effectively embed the DNN-based object detector into an edge device equipped with Qualcomm’s QCS605 System-on-Chip (SoC), while achieving a real-time operation with more than 10 frames per second.

Schlagwörter

Stichwörter der Autor:innen

channel pruningedge computingnetwork simplificationobject detector

Autor:innen und Institutionen

  • Department of AI Automation Robot, Daegu Catholic University, 13-13 Hayang-ro, Hayang-eup, Gyeongsan-si 38430, Gyeongsangbuk-do, Republic of Korea.
  • Driving Image Recognition Logic Cell, Hyundai Mobis, 17-2 Mabuk-ro 240beon-gil, Giheung-gu, Yongin-si 16891, Gyeonggi-do, Republic of Korea.
  • Department of Electronic Engineering, Korea National University of Transportation, 50 Daehak-ro, Chungju-si 27469, Chungbuk-do, Republic of Korea.
  • Department of Intelligent Mechatronics Engineering, Sejong University, 209 Neungdong-ro, Gwangjin-gu, Seoul 05006, Republic of Korea.

ORCID-Profile: Kyoungtaek Choi, Ho Gi Jung, Jae Kyu Suhr

Förderung

  • National Research Foundation of Korea
  • Ministry of Science ICT and Future Planning

Zitieren

APA

Choi, K., Wi, S. M., Jung, H. G., & Suhr, J. K. (2023). Simplification of Deep Neural Network-Based Object Detector for Real-Time Edge Computing. Sensors, 23(7), 3777. https://doi.org/10.3390/s23073777

Vancouver

Choi K, Wi SM, Jung HG, Suhr JK. Simplification of Deep Neural Network-Based Object Detector for Real-Time Edge Computing. Sensors. 2023;23(7):3777. doi:10.3390/s23073777

BibTeX
@article{Choi2023,
  author = {Choi, Kyoungtaek and Wi, Seong Min and Jung, Ho Gi and Suhr, Jae Kyu},
  title = {Simplification of Deep Neural Network-Based Object Detector for Real-Time Edge Computing},
  journal = {Sensors},
  year = {2023},
  volume = {23},
  number = {7},
  pages = {3777},
  doi = {10.3390/s23073777}
}
Kostenlos starten →