Improving a real-time object detector with compact temporal information

Martin Ahrnbom, Morten Bornø Jensen, Kalle Åström, Mikael Nilsson, Håkan Ardö, Thomas B. Moeslund

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

234 Downloads (Pure)

Abstract

Neural networks designed for real-time object detection have recently improved significantly, but in practice, look- ing at only a single RGB image at the time may not be ideal. For example, when detecting objects in videos, a foreground detection algorithm can be used to obtain compact temporal data, which can be fed into a neural network alongside RGB images. We propose an approach for doing this, based on an existing object detector, that re-uses pretrained weights for the processing of RGB images. The neural network was tested on the VIRAT dataset with annotations for object de- tection, a problem this approach is well suited for. The ac- curacy was found to improve significantly (up to 66%), with a roughly 40% increase in computational time.
Original languageEnglish
Title of host publication2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop
Number of pages8
PublisherIEEE
Publication date2017
Pages190-197
ISBN (Electronic)978-1-5386-1034-3
DOIs
Publication statusPublished - 2017
Event2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop - PALAZZO DEL CINEMA - VENICE CONVENTION CENTER, Venice, Italy
Duration: 22 Oct 201729 Oct 2017

Conference

Conference2017 IEEE International Conference on Computer Vision Workshop (ICCVW)
LocationPALAZZO DEL CINEMA - VENICE CONVENTION CENTER
CountryItaly
CityVenice
Period22/10/201729/10/2017
SeriesIEEE International Conference on Computer Vision Workshops (ICCVW)
ISSN2473-9944

Fingerprint

Detectors
Neural networks
Processing
Object detection

Cite this

Ahrnbom, M., Jensen, M. B., Åström, K., Nilsson, M., Ardö, H., & Moeslund, T. B. (2017). Improving a real-time object detector with compact temporal information. In 2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop (pp. 190-197). IEEE. IEEE International Conference on Computer Vision Workshops (ICCVW) https://doi.org/10.1109/ICCVW.2017.31
Ahrnbom, Martin ; Jensen, Morten Bornø ; Åström, Kalle ; Nilsson, Mikael ; Ardö, Håkan ; Moeslund, Thomas B. / Improving a real-time object detector with compact temporal information. 2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop. IEEE, 2017. pp. 190-197 (IEEE International Conference on Computer Vision Workshops (ICCVW)).
@inproceedings{5c344fce3d8e4bad90ed93bff1a15ef6,
title = "Improving a real-time object detector with compact temporal information",
abstract = "Neural networks designed for real-time object detection have recently improved significantly, but in practice, look- ing at only a single RGB image at the time may not be ideal. For example, when detecting objects in videos, a foreground detection algorithm can be used to obtain compact temporal data, which can be fed into a neural network alongside RGB images. We propose an approach for doing this, based on an existing object detector, that re-uses pretrained weights for the processing of RGB images. The neural network was tested on the VIRAT dataset with annotations for object de- tection, a problem this approach is well suited for. The ac- curacy was found to improve significantly (up to 66{\%}), with a roughly 40{\%} increase in computational time.",
author = "Martin Ahrnbom and Jensen, {Morten Born{\o}} and Kalle {\AA}str{\"o}m and Mikael Nilsson and H{\aa}kan Ard{\"o} and Moeslund, {Thomas B.}",
year = "2017",
doi = "10.1109/ICCVW.2017.31",
language = "English",
series = "IEEE International Conference on Computer Vision Workshops (ICCVW)",
publisher = "IEEE",
pages = "190--197",
booktitle = "2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop",
address = "United States",

}

Ahrnbom, M, Jensen, MB, Åström, K, Nilsson, M, Ardö, H & Moeslund, TB 2017, Improving a real-time object detector with compact temporal information. in 2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop. IEEE, IEEE International Conference on Computer Vision Workshops (ICCVW), pp. 190-197, Venice, Italy, 22/10/2017. https://doi.org/10.1109/ICCVW.2017.31

Improving a real-time object detector with compact temporal information. / Ahrnbom, Martin; Jensen, Morten Bornø; Åström, Kalle; Nilsson, Mikael; Ardö, Håkan; Moeslund, Thomas B.

2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop. IEEE, 2017. p. 190-197 (IEEE International Conference on Computer Vision Workshops (ICCVW)).

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Improving a real-time object detector with compact temporal information

AU - Ahrnbom, Martin

AU - Jensen, Morten Bornø

AU - Åström, Kalle

AU - Nilsson, Mikael

AU - Ardö, Håkan

AU - Moeslund, Thomas B.

PY - 2017

Y1 - 2017

N2 - Neural networks designed for real-time object detection have recently improved significantly, but in practice, look- ing at only a single RGB image at the time may not be ideal. For example, when detecting objects in videos, a foreground detection algorithm can be used to obtain compact temporal data, which can be fed into a neural network alongside RGB images. We propose an approach for doing this, based on an existing object detector, that re-uses pretrained weights for the processing of RGB images. The neural network was tested on the VIRAT dataset with annotations for object de- tection, a problem this approach is well suited for. The ac- curacy was found to improve significantly (up to 66%), with a roughly 40% increase in computational time.

AB - Neural networks designed for real-time object detection have recently improved significantly, but in practice, look- ing at only a single RGB image at the time may not be ideal. For example, when detecting objects in videos, a foreground detection algorithm can be used to obtain compact temporal data, which can be fed into a neural network alongside RGB images. We propose an approach for doing this, based on an existing object detector, that re-uses pretrained weights for the processing of RGB images. The neural network was tested on the VIRAT dataset with annotations for object de- tection, a problem this approach is well suited for. The ac- curacy was found to improve significantly (up to 66%), with a roughly 40% increase in computational time.

U2 - 10.1109/ICCVW.2017.31

DO - 10.1109/ICCVW.2017.31

M3 - Article in proceeding

T3 - IEEE International Conference on Computer Vision Workshops (ICCVW)

SP - 190

EP - 197

BT - 2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop

PB - IEEE

ER -

Ahrnbom M, Jensen MB, Åström K, Nilsson M, Ardö H, Moeslund TB. Improving a real-time object detector with compact temporal information. In 2017 IEEE International Conference on Computer Vision Workshop (ICCVW): Computer Vision for Road Scene Understanding and Autonomous Driving workshop. IEEE. 2017. p. 190-197. (IEEE International Conference on Computer Vision Workshops (ICCVW)). https://doi.org/10.1109/ICCVW.2017.31