Ghent University Academic Bibliography

Advanced

Information feedback loop for improved pedestrian detection in an autonomous perception system

Martin Dimitrievski UGent, Peter Veelaert UGent and Wilfried Philips UGent (2018) In The 21st IEEE International Conference on Intelligent Transportation Systems ITSC
abstract
Environmental perception systems for autonomous vehicles are often built using heterogeneous technologies that operate in a sequential manner. In the task of object tracking in particular, where the classical detector-tracker interaction is a serial process, it is viable to break the design rule by introducing information loops. This is especially feasible in a tracker that operates in a prediction-update cycle. Tracking predictions can steer object detection towards regions where an object is anticipated and, in turn, tracking updates can be improved by incorporating reinforced detections. In this paper we propose a novel detector-tracker feedback loop for information exchange based on the spatio-temporal similarity of detections and tracklets. We reinforce pedestrian detections that have weak confidence scores by matching their bounding boxes to estimated tracklets with high tracking confidence. The proposed system has several compelling advantages: based on a positive feedback principle it extracts the maximum detection and tracking information, while operating transparently and with minimal computational load. In a controlled ablation study we evaluate our feedback mechanism using the KITTI object tracking dataset. We show that our system gains significant performance increase over the baseline in both frame-by-frame detection and tracking quality.
Please use this url to cite or link to this publication:
author
organization
year
type
conference (proceedingsPaper)
publication status
in press
subject
keyword
pedestrian detection, object detection, deep learning, tracking, multi-object tracking, feedback loop, autonomous vehicles, environmental perception
series title
The 21st IEEE International Conference on Intelligent Transportation Systems ITSC
conference name
The 21st IEEE International Conference on Intelligent Transportation Systems
conference organizer
IEEE Intelligent Transportation Systems Society
conference location
Maui, Hawaii, USA
conference start
2018-11-04
conference end
2018-11-7
language
English
UGent publication?
yes
classification
U
copyright statement
I don't know the status of the copyright for this publication
id
8567675
handle
http://hdl.handle.net/1854/LU-8567675
date created
2018-07-02 14:30:35
date last changed
2018-07-04 08:19:24
@inproceedings{8567675,
  abstract     = {Environmental perception systems for autonomous
vehicles are often built using heterogeneous technologies that
operate in a sequential manner. In the task of object tracking
in particular, where the classical detector-tracker interaction
is a serial process, it is viable to break the design rule by
introducing information loops. This is especially feasible in a
tracker that operates in a prediction-update cycle. Tracking
predictions can steer object detection towards regions where
an object is anticipated and, in turn, tracking updates can
be improved by incorporating reinforced detections. In this
paper we propose a novel detector-tracker feedback loop for
information exchange based on the spatio-temporal similarity
of detections and tracklets. We reinforce pedestrian detections
that have weak confidence scores by matching their bounding
boxes to estimated tracklets with high tracking confidence. The
proposed system has several compelling advantages: based on
a positive feedback principle it extracts the maximum detection
and tracking information, while operating transparently and
with minimal computational load. In a controlled ablation study
we evaluate our feedback mechanism using the KITTI object
tracking dataset. We show that our system gains significant
performance increase over the baseline in both frame-by-frame
detection and tracking quality.},
  author       = {Dimitrievski, Martin and Veelaert, Peter and Philips, Wilfried},
  keyword      = {pedestrian detection,object detection,deep learning,tracking,multi-object tracking,feedback loop,autonomous vehicles,environmental perception},
  language     = {eng},
  location     = {Maui, Hawaii, USA},
  title        = {Information feedback loop for improved pedestrian detection in an autonomous perception system},
  year         = {2018},
}

Chicago
Dimitrievski, Martin, Peter Veelaert, and Wilfried Philips. 2018. “Information Feedback Loop for Improved Pedestrian Detection in an Autonomous Perception System.” In .
APA
Dimitrievski, M., Veelaert, P., & Philips, W. (2018). Information feedback loop for improved pedestrian detection in an autonomous perception system. Presented at the The 21st IEEE International Conference on Intelligent Transportation Systems.
Vancouver
1.
Dimitrievski M, Veelaert P, Philips W. Information feedback loop for improved pedestrian detection in an autonomous perception system. 2018.
MLA
Dimitrievski, Martin, Peter Veelaert, and Wilfried Philips. “Information Feedback Loop for Improved Pedestrian Detection in an Autonomous Perception System.” 2018. Print.