Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Seg2Pose: Pose Estimations from Instance Segmentation Masks in One or Multiple Views for Traffic Applications

Ahrnbom, Martin LU orcid ; Persson, Ivar LU and Nilsson, Mikael LU (2022) 17th International Conference on Computer Vision Theory and Applications, VISAPP 2022
5. p.777-784
Abstract
A system we denote Seg2Pose is presented which converts pixel coordinate tracks, represented by instance segmentation masks across multiple video frames, into world coordinate pose tracks, for road users seen by static surveillance cameras. The road users are bound to a ground surface represented by a number of 3D points and does not necessarily have to be perfectly flat. The system works with one or more views, by using a late fusion scheme. An approximate position, denoted the normal position, is computed from the camera calibration, per-class default heights and the ground surface model. The position is then refined a novel Convolutional Neural Network we denote Seg2PoseNet, taking instance segmentations and cropping positioning as its... (More)
A system we denote Seg2Pose is presented which converts pixel coordinate tracks, represented by instance segmentation masks across multiple video frames, into world coordinate pose tracks, for road users seen by static surveillance cameras. The road users are bound to a ground surface represented by a number of 3D points and does not necessarily have to be perfectly flat. The system works with one or more views, by using a late fusion scheme. An approximate position, denoted the normal position, is computed from the camera calibration, per-class default heights and the ground surface model. The position is then refined a novel Convolutional Neural Network we denote Seg2PoseNet, taking instance segmentations and cropping positioning as its input. We evaluate this system quantitatively both on synthetic data from CARLA Simulator and on a real recording from a trinocular camera. The system outperforms the baseline method of only using the normal positions, which is roughly equivalent of a typical 2D to 3D conversion system, in both datasets. (Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
keywords
Pose Estimation, Instance Segmentation, Convolutional Neural Network, Traffic Safety, Road Users, Tracking, Stereo Camera, Trinocular Camera Array, Traffic Surveillance
host publication
Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications: VISAPP
volume
5
pages
8 pages
publisher
SciTePress
conference name
17th International Conference on Computer Vision Theory and Applications, VISAPP 2022<br/>
conference location
Virtual, Online Streaming
conference dates
2022-02-06 - 2022-02-08
external identifiers
  • scopus:85143899640
ISBN
978-989-758-555-5
DOI
10.5220/0010777700003124
language
Swedish
LU publication?
yes
id
758b9f34-9b6a-49a3-b0c9-b78118b1b442
date added to LUP
2022-02-16 13:13:32
date last changed
2024-04-06 04:02:18
@inproceedings{758b9f34-9b6a-49a3-b0c9-b78118b1b442,
  abstract     = {{A system we denote Seg2Pose is presented which converts pixel coordinate tracks, represented by instance segmentation masks across multiple video frames, into world coordinate pose tracks, for road users seen by static surveillance cameras. The road users are bound to a ground surface represented by a number of 3D points and does not necessarily have to be perfectly flat. The system works with one or more views, by using a late fusion scheme. An approximate position, denoted the normal position, is computed from the camera calibration, per-class default heights and the ground surface model. The position is then refined a novel Convolutional Neural Network we denote Seg2PoseNet, taking instance segmentations and cropping positioning as its input. We evaluate this system quantitatively both on synthetic data from CARLA Simulator and on a real recording from a trinocular camera. The system outperforms the baseline method of only using the normal positions, which is roughly equivalent of a typical 2D to 3D conversion system, in both datasets.}},
  author       = {{Ahrnbom, Martin and Persson, Ivar and Nilsson, Mikael}},
  booktitle    = {{Proceedings of the 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications: VISAPP}},
  isbn         = {{978-989-758-555-5}},
  keywords     = {{Pose Estimation; Instance Segmentation; Convolutional Neural Network; Traffic Safety; Road Users; Tracking; Stereo Camera; Trinocular Camera Array; Traffic Surveillance}},
  language     = {{swe}},
  month        = {{02}},
  pages        = {{777--784}},
  publisher    = {{SciTePress}},
  title        = {{Seg2Pose: Pose Estimations from Instance Segmentation Masks in One or Multiple Views for Traffic Applications}},
  url          = {{http://dx.doi.org/10.5220/0010777700003124}},
  doi          = {{10.5220/0010777700003124}},
  volume       = {{5}},
  year         = {{2022}},
}