Near-chip dynamic vision filtering for low-bandwidth pedestrian detection

Anthony Bisulco; Fernando Cladera Ojeda; Volkan Isler; Daniel Dongyuel Lee

doi:10.1109/ISVLSI49217.2020.00050

Near-chip dynamic vision filtering for low-bandwidth pedestrian detection

Anthony Bisulco, Fernando Cladera Ojeda, Volkan Isler, Daniel Dongyuel Lee

Computer Science and Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

3 Scopus citations

Abstract

This paper presents a novel end-to-end system for pedestrian detection using Dynamic Vision Sensors (DVSs). We target applications where multiple sensors transmit data to a local processing unit, which executes a detection algorithm. Our system is composed of (i) a near-chip event filter that compresses and denoises the event stream from the DVS, and (ii) a Binary Neural Network (BNN) detection module that runs on a low-computation edge computing device (in our case a STM32F4 microcontroller). We present the system architecture and provide an end-to-end implementation for pedestrian detection in an office environment. Our implementation reduces transmission size by up to 99.6% compared to transmitting the raw event stream. Our detector is able to perform a detection every 450 ms, with an overall testing F1 score of 83%. The low bandwidth and energy properties of our system make it ideal for IoT applications.

Original language	English (US)
Title of host publication	Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020
Publisher	IEEE Computer Society
Pages	234-239
Number of pages	6
ISBN (Electronic)	9781728157757
DOIs	https://doi.org/10.1109/ISVLSI49217.2020.00050
State	Published - Jul 2020
Event	19th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020 - Limassol, Cyprus Duration: Jul 6 2020 → Jul 8 2020

Publication series

Name	Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI
Volume	2020-July
ISSN (Print)	2159-3469
ISSN (Electronic)	2159-3477

Conference

Conference	19th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020
Country/Territory	Cyprus
City	Limassol
Period	7/6/20 → 7/8/20

Bibliographical note

Publisher Copyright:
© 2020 IEEE.

Keywords

Binary neural networks
Dynamic vision sensors
FPGA
Pedestrian detection

Publisher link

10.1109/ISVLSI49217.2020.00050

Find It

Find It at U of M Libraries

Cite this

Bisulco, A., Cladera Ojeda, F., Isler, V., & Lee, D. D. (2020). Near-chip dynamic vision filtering for low-bandwidth pedestrian detection. In Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020 (pp. 234-239). Article 9155035 (Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI; Vol. 2020-July). IEEE Computer Society. https://doi.org/10.1109/ISVLSI49217.2020.00050

Near-chip dynamic vision filtering for low-bandwidth pedestrian detection. / Bisulco, Anthony; Cladera Ojeda, Fernando; Isler, Volkan et al.
Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020. IEEE Computer Society, 2020. p. 234-239 9155035 (Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI; Vol. 2020-July).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Bisulco, A, Cladera Ojeda, F, Isler, V & Lee, DD 2020, Near-chip dynamic vision filtering for low-bandwidth pedestrian detection. in Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020., 9155035, Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI, vol. 2020-July, IEEE Computer Society, pp. 234-239, 19th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020, Limassol, Cyprus, 7/6/20. https://doi.org/10.1109/ISVLSI49217.2020.00050

@inproceedings{115c860c7ab54db8a9f9c8b9c717492a,

title = "Near-chip dynamic vision filtering for low-bandwidth pedestrian detection",

abstract = "This paper presents a novel end-to-end system for pedestrian detection using Dynamic Vision Sensors (DVSs). We target applications where multiple sensors transmit data to a local processing unit, which executes a detection algorithm. Our system is composed of (i) a near-chip event filter that compresses and denoises the event stream from the DVS, and (ii) a Binary Neural Network (BNN) detection module that runs on a low-computation edge computing device (in our case a STM32F4 microcontroller). We present the system architecture and provide an end-to-end implementation for pedestrian detection in an office environment. Our implementation reduces transmission size by up to 99.6% compared to transmitting the raw event stream. Our detector is able to perform a detection every 450 ms, with an overall testing F1 score of 83%. The low bandwidth and energy properties of our system make it ideal for IoT applications.",

keywords = "Binary neural networks, Dynamic vision sensors, FPGA, Pedestrian detection",

author = "Anthony Bisulco and {Cladera Ojeda}, Fernando and Volkan Isler and Lee, {Daniel Dongyuel}",

note = "Publisher Copyright: {\textcopyright} 2020 IEEE.; 19th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020 ; Conference date: 06-07-2020 Through 08-07-2020",

year = "2020",

month = jul,

doi = "10.1109/ISVLSI49217.2020.00050",

language = "English (US)",

series = "Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI",

publisher = "IEEE Computer Society",

pages = "234--239",

booktitle = "Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020",

}

TY - GEN

T1 - Near-chip dynamic vision filtering for low-bandwidth pedestrian detection

AU - Bisulco, Anthony

AU - Cladera Ojeda, Fernando

AU - Isler, Volkan

AU - Lee, Daniel Dongyuel

PY - 2020/7

Y1 - 2020/7

N2 - This paper presents a novel end-to-end system for pedestrian detection using Dynamic Vision Sensors (DVSs). We target applications where multiple sensors transmit data to a local processing unit, which executes a detection algorithm. Our system is composed of (i) a near-chip event filter that compresses and denoises the event stream from the DVS, and (ii) a Binary Neural Network (BNN) detection module that runs on a low-computation edge computing device (in our case a STM32F4 microcontroller). We present the system architecture and provide an end-to-end implementation for pedestrian detection in an office environment. Our implementation reduces transmission size by up to 99.6% compared to transmitting the raw event stream. Our detector is able to perform a detection every 450 ms, with an overall testing F1 score of 83%. The low bandwidth and energy properties of our system make it ideal for IoT applications.

AB - This paper presents a novel end-to-end system for pedestrian detection using Dynamic Vision Sensors (DVSs). We target applications where multiple sensors transmit data to a local processing unit, which executes a detection algorithm. Our system is composed of (i) a near-chip event filter that compresses and denoises the event stream from the DVS, and (ii) a Binary Neural Network (BNN) detection module that runs on a low-computation edge computing device (in our case a STM32F4 microcontroller). We present the system architecture and provide an end-to-end implementation for pedestrian detection in an office environment. Our implementation reduces transmission size by up to 99.6% compared to transmitting the raw event stream. Our detector is able to perform a detection every 450 ms, with an overall testing F1 score of 83%. The low bandwidth and energy properties of our system make it ideal for IoT applications.

KW - Binary neural networks

KW - Dynamic vision sensors

KW - FPGA

KW - Pedestrian detection

UR - http://www.scopus.com/inward/record.url?scp=85090420913&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85090420913&partnerID=8YFLogxK

U2 - 10.1109/ISVLSI49217.2020.00050

DO - 10.1109/ISVLSI49217.2020.00050

M3 - Conference contribution

AN - SCOPUS:85090420913

T3 - Proceedings of IEEE Computer Society Annual Symposium on VLSI, ISVLSI

SP - 234

EP - 239

BT - Proceedings - 2020 IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020

PB - IEEE Computer Society

T2 - 19th IEEE Computer Society Annual Symposium on VLSI, ISVLSI 2020

Y2 - 6 July 2020 through 8 July 2020

ER -

Near-chip dynamic vision filtering for low-bandwidth pedestrian detection

Abstract

Publication series

Conference

Bibliographical note

Keywords

Publisher link

Find It

Other files and links

Fingerprint

Cite this