RetinaNet Object Detector based on Analog-to-Spiking Neural Network  Conversion

Joaquin Royo-Miquel; Silvia Tolu; Frederik E. T. Schöller; Roberto Galeazzi

doi:10.1109/ISCMI53840.2021.9654818

RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion

Joaquin Royo-Miquel, Silvia Tolu, Frederik E. T. Schöller, Roberto Galeazzi

Technical University of Denmark

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

175 Downloads (Pure)

Abstract

The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.

Original language	English
Title of host publication	Proceedings of 2021 8^th International Conference on Soft Computing & Machine Intelligence
Number of pages	5
Publisher	IEEE
Publication date	2021
ISBN (Print)	978-1-7281-8684-9
DOIs	https://doi.org/10.1109/ISCMI53840.2021.9654818
Publication status	Published - 2021
Event	2021 8^th International Conference on Soft Computing & Machine Intelligence - Steigenberger Hotel El Tahrir, Cairo, Egypt Duration: 26 Nov 2021 → 27 Nov 2021 Conference number: 8 http://www.iscmi.us

Conference

Conference	2021 8^th International Conference on Soft Computing & Machine Intelligence
Number	8
Location	Steigenberger Hotel El Tahrir
Country/Territory	Egypt
City	Cairo
Period	26/11/2021 → 27/11/2021
Internet address	http://www.iscmi.us

Keywords

Spiking Neural Networks
Object Detection
Spiking-RetinaNet

Access to Document

10.1109/ISCMI53840.2021.9654818

FulltextAccepted author manuscript, 2.1 MB

OpenUrl availability

Full text

Cite this

@inproceedings{098185c7a3ea419c93919895c76575b9,

title = "RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion",

abstract = "The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.",

keywords = "Spiking Neural Networks, Object Detection, Spiking-RetinaNet",

author = "Joaquin Royo-Miquel and Silvia Tolu and Sch{\"o}ller, {Frederik E. T.} and Roberto Galeazzi",

year = "2021",

doi = "10.1109/ISCMI53840.2021.9654818",

language = "English",

isbn = "978-1-7281-8684-9",

booktitle = "Proceedings of 2021 8th International Conference on Soft Computing & Machine Intelligence",

publisher = "IEEE",

address = "United States",

note = "2021 8<sup>th</sup> International Conference on Soft Computing & Machine Intelligence, ISCMI2021 ; Conference date: 26-11-2021 Through 27-11-2021",

url = "http://www.iscmi.us",

}

Royo-Miquel, J, Tolu, S, Schöller, FET & Galeazzi, R 2021, RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion. in Proceedings of 2021 8^th International Conference on Soft Computing & Machine Intelligence. IEEE, 2021 8^th International Conference on Soft Computing & Machine Intelligence, Cairo, Egypt, 26/11/2021. https://doi.org/10.1109/ISCMI53840.2021.9654818

RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion. / Royo-Miquel, Joaquin; Tolu, Silvia; Schöller, Frederik E. T. et al.
Proceedings of 2021 8^th International Conference on Soft Computing & Machine Intelligence. IEEE, 2021.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - RetinaNet Object Detector based on Analog-to-Spiking Neural Network Conversion

AU - Royo-Miquel, Joaquin

AU - Tolu, Silvia

AU - Schöller, Frederik E. T.

AU - Galeazzi, Roberto

N1 - Conference code: 8

PY - 2021

Y1 - 2021

N2 - The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.

AB - The paper proposes a method to convert a deep learning object detector into an equivalent spiking neural network. The aim is to provide a conversion framework that is not constrained to shallow network structures and classification problems as in state-of-the-art conversion libraries. The results show that models of higher complexity, such as the RetinaNet object detector, can be converted with limited loss in performance.

KW - Spiking Neural Networks

KW - Object Detection

KW - Spiking-RetinaNet

U2 - 10.1109/ISCMI53840.2021.9654818

DO - 10.1109/ISCMI53840.2021.9654818

M3 - Article in proceedings

SN - 978-1-7281-8684-9

BT - Proceedings of 2021 8th International Conference on Soft Computing & Machine Intelligence

PB - IEEE

T2 - 2021 8<sup>th</sup> International Conference on Soft Computing & Machine Intelligence

Y2 - 26 November 2021 through 27 November 2021

ER -