VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN

Research output: Contribution to journalConference articleResearchpeer review

Authors

  • S. El Amrani Abouelassad
  • F. Rottensteiner

Details

Original languageEnglish
Pages (from-to)15-23
Number of pages9
JournalISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Volume5
Issue number1
Publication statusPublished - 17 May 2022
Event2022 24th ISPRS Congress on Imaging Today, Foreseeing Tomorrow, Commission I - Nice, France
Duration: 6 Jun 202211 Jun 2022

Abstract

Vehicle instance segmentation is a major but challenging task in aerial remote sensing applications. More importantly, the current majority methods use horizontal bounding boxes which does not tell much about the orientation of vehicles and often leads to inaccurate mask proposals due to high background to foreground pixel-ratio. Given that the orientation of vehicles is important for numerous applications like vehicle tracking, we introduce in this paper a deep neural network to detect and segment vehicles using rotated bounding boxes in aerial images. Our method demonstrates that rotated instance segmentation improves the mask predictions, especially when objects are not axis aligned or are touching. We evaluate our model on the ISPRS benchmark dataset and our newly introduced UAV dataset for vehicle segmentation and show that we can significantly improve the mask accuracy compared to instance segmentation using axis-aligned bounding boxes.

Keywords

    Deep Learning, Instance Segmentation, Mask RCNN, Remote Sensing, Vehicle Detection

ASJC Scopus subject areas

Cite this

VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN. / El Amrani Abouelassad, S.; Rottensteiner, F.
In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 5, No. 1, 17.05.2022, p. 15-23.

Research output: Contribution to journalConference articleResearchpeer review

El Amrani Abouelassad, S & Rottensteiner, F 2022, 'VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN', ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 5, no. 1, pp. 15-23. https://doi.org/10.5194/isprs-annals-V-1-2022-15-2022
El Amrani Abouelassad, S., & Rottensteiner, F. (2022). VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 5(1), 15-23. https://doi.org/10.5194/isprs-annals-V-1-2022-15-2022
El Amrani Abouelassad S, Rottensteiner F. VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2022 May 17;5(1):15-23. doi: 10.5194/isprs-annals-V-1-2022-15-2022
El Amrani Abouelassad, S. ; Rottensteiner, F. / VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN. In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. 2022 ; Vol. 5, No. 1. pp. 15-23.
Download
@article{5683e9fc2cd042908f1f4ca3fc307955,
title = "VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN",
abstract = "Vehicle instance segmentation is a major but challenging task in aerial remote sensing applications. More importantly, the current majority methods use horizontal bounding boxes which does not tell much about the orientation of vehicles and often leads to inaccurate mask proposals due to high background to foreground pixel-ratio. Given that the orientation of vehicles is important for numerous applications like vehicle tracking, we introduce in this paper a deep neural network to detect and segment vehicles using rotated bounding boxes in aerial images. Our method demonstrates that rotated instance segmentation improves the mask predictions, especially when objects are not axis aligned or are touching. We evaluate our model on the ISPRS benchmark dataset and our newly introduced UAV dataset for vehicle segmentation and show that we can significantly improve the mask accuracy compared to instance segmentation using axis-aligned bounding boxes.",
keywords = "Deep Learning, Instance Segmentation, Mask RCNN, Remote Sensing, Vehicle Detection",
author = "{El Amrani Abouelassad}, S. and F. Rottensteiner",
note = "Funding Information: This work was supported by the German Research Foundation (DFG) as a part of the Research Training Group i.c.sens [GRK2159].; 2022 24th ISPRS Congress on Imaging Today, Foreseeing Tomorrow, Commission I ; Conference date: 06-06-2022 Through 11-06-2022",
year = "2022",
month = may,
day = "17",
doi = "10.5194/isprs-annals-V-1-2022-15-2022",
language = "English",
volume = "5",
pages = "15--23",
number = "1",

}

Download

TY - JOUR

T1 - VEHICLE INSTANCE SEGMENTATION WITH ROTATED BOUNDING BOXES IN UAV IMAGES USING CNN

AU - El Amrani Abouelassad, S.

AU - Rottensteiner, F.

N1 - Funding Information: This work was supported by the German Research Foundation (DFG) as a part of the Research Training Group i.c.sens [GRK2159].

PY - 2022/5/17

Y1 - 2022/5/17

N2 - Vehicle instance segmentation is a major but challenging task in aerial remote sensing applications. More importantly, the current majority methods use horizontal bounding boxes which does not tell much about the orientation of vehicles and often leads to inaccurate mask proposals due to high background to foreground pixel-ratio. Given that the orientation of vehicles is important for numerous applications like vehicle tracking, we introduce in this paper a deep neural network to detect and segment vehicles using rotated bounding boxes in aerial images. Our method demonstrates that rotated instance segmentation improves the mask predictions, especially when objects are not axis aligned or are touching. We evaluate our model on the ISPRS benchmark dataset and our newly introduced UAV dataset for vehicle segmentation and show that we can significantly improve the mask accuracy compared to instance segmentation using axis-aligned bounding boxes.

AB - Vehicle instance segmentation is a major but challenging task in aerial remote sensing applications. More importantly, the current majority methods use horizontal bounding boxes which does not tell much about the orientation of vehicles and often leads to inaccurate mask proposals due to high background to foreground pixel-ratio. Given that the orientation of vehicles is important for numerous applications like vehicle tracking, we introduce in this paper a deep neural network to detect and segment vehicles using rotated bounding boxes in aerial images. Our method demonstrates that rotated instance segmentation improves the mask predictions, especially when objects are not axis aligned or are touching. We evaluate our model on the ISPRS benchmark dataset and our newly introduced UAV dataset for vehicle segmentation and show that we can significantly improve the mask accuracy compared to instance segmentation using axis-aligned bounding boxes.

KW - Deep Learning

KW - Instance Segmentation

KW - Mask RCNN

KW - Remote Sensing

KW - Vehicle Detection

UR - http://www.scopus.com/inward/record.url?scp=85132795615&partnerID=8YFLogxK

U2 - 10.5194/isprs-annals-V-1-2022-15-2022

DO - 10.5194/isprs-annals-V-1-2022-15-2022

M3 - Conference article

AN - SCOPUS:85132795615

VL - 5

SP - 15

EP - 23

JO - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

JF - ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences

SN - 2194-9042

IS - 1

T2 - 2022 24th ISPRS Congress on Imaging Today, Foreseeing Tomorrow, Commission I

Y2 - 6 June 2022 through 11 June 2022

ER -