Advanced

Improving a real-time object detector with compact temporal information

Ahrnbom, Martin LU ; Bornø Jensen, Morten; Åström, Karl LU ; Nilsson, Mikael LU ; Ardö, Håkan LU and Moeslund, Thomas (2018) 16th IEEE International Conference on Computer Vision Workshops, ICCVW 2017 In International Conference on Computer Vision Workshops, 2017 p.190-197
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%),... (More)
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. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
in
International Conference on Computer Vision Workshops, 2017
pages
8 pages
publisher
IEEE--Institute of Electrical and Electronics Engineers Inc.
conference name
16th IEEE International Conference on Computer Vision Workshops, ICCVW 2017
external identifiers
  • scopus:85046282569
DOI
10.1109/ICCVW.2017.31
language
English
LU publication?
yes
id
41030db4-c01e-4609-be48-63157ea594d4
alternative location
http://openaccess.thecvf.com/content_ICCV_2017_workshops/papers/w3/Ahrnbom_Improving_a_Real-Time_ICCV_2017_paper.pdf
date added to LUP
2018-01-27 08:51:36
date last changed
2018-05-20 04:40:22
@inproceedings{41030db4-c01e-4609-be48-63157ea594d4,
  abstract     = {Neural networks designed for real-time object detection<br/>have recently improved significantly, but in practice, look-<br/>ing at only a single RGB image at the time may not be ideal.<br/>For example, when detecting objects in videos, a foreground<br/>detection algorithm can be used to obtain compact temporal<br/>data, which can be fed into a neural network alongside RGB<br/>images. We propose an approach for doing this, based on<br/>an existing object detector, that re-uses pretrained weights<br/>for the processing of RGB images. The neural network was<br/>tested on the VIRAT dataset with annotations for object de-<br/>tection, a problem this approach is well suited for. The ac-<br/>curacy was found to improve significantly (up to 66%), with<br/>a roughly 40% increase in computational time.},
  author       = {Ahrnbom, Martin and Bornø Jensen, Morten and Åström, Karl and Nilsson, Mikael and Ardö, Håkan and Moeslund, Thomas},
  booktitle    = {International Conference on Computer Vision Workshops, 2017},
  language     = {eng},
  month        = {01},
  pages        = {190--197},
  publisher    = {IEEE--Institute of Electrical and Electronics Engineers Inc.},
  title        = {Improving a real-time object detector with compact temporal information},
  url          = {http://dx.doi.org/10.1109/ICCVW.2017.31},
  year         = {2018},
}