Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task

Jyri Maanpää; Iaroslav Melekhov; Josef Taher; Petri Manninen; Juha Hyyppä

doi:10.1007/978-3-031-06427-2_60

Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task

Jyri Maanpää^*, Iaroslav Melekhov, Josef Taher, Petri Manninen, Juha Hyyppä

^*Corresponding author for this work

National Land Survey of Finland

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

1 Citation (Scopus)

Abstract

Robustness of different pattern recognition methods is one of the key challenges in autonomous driving, especially when driving in the high variety of road environments and weather conditions, such as gravel roads and snowfall. Although one can collect data from these adverse conditions using cars equipped with sensors, it is quite tedious to annotate the data for training. In this work, we address this limitation and propose a CNN-based method that can leverage the steering wheel angle information to improve the road area semantic segmentation. As the steering wheel angle data can be easily acquired with the associated images, one could improve the accuracy of road area semantic segmentation by collecting data in new road environments without manual data annotation. We demonstrate the effectiveness of the proposed approach on two challenging data sets for autonomous driving and show that when the steering task is used in our segmentation model training, it leads to a 0.1–2.9% gain in the road area mIoU (mean Intersection over Union) compared to the corresponding reference transfer learning model.

Original language	English
Title of host publication	Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings
Editors	Stan Sclaroff, Cosimo Distante, Marco Leo, Giovanni M. Farinella, Federico Tombari
Publisher	Springer
Pages	727-738
Number of pages	12
ISBN (Print)	978-3-031-06426-5
DOIs	https://doi.org/10.1007/978-3-031-06427-2_60
Publication status	Published - 2022
MoE publication type	A4 Conference publication
Event	International Conference on Image Analysis and Processing - Lecce, Italy Duration: 23 May 2022 → 27 May 2022 Conference number: 21

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer
Volume	13231 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International Conference on Image Analysis and Processing
Abbreviated title	ICIAP
Country/Territory	Italy
City	Lecce
Period	23/05/2022 → 27/05/2022

Keywords

Autonomous driving
Domain adaptation
Multi-task learning
Road area semantic segmentation
Transfer learning

Access to Document

10.1007/978-3-031-06427-2_60

https://arxiv.org/abs/2212.09438

Cite this

Maanpää, J., Melekhov, I., Taher, J., Manninen, P., & Hyyppä, J. (2022). Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task. In S. Sclaroff, C. Distante, M. Leo, G. M. Farinella, & F. Tombari (Eds.), Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings (pp. 727-738). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13231 LNCS). Springer. https://doi.org/10.1007/978-3-031-06427-2_60

Maanpää, Jyri ; Melekhov, Iaroslav ; Taher, Josef et al. / Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task. Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings. editor / Stan Sclaroff ; Cosimo Distante ; Marco Leo ; Giovanni M. Farinella ; Federico Tombari. Springer, 2022. pp. 727-738 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{699a64ce9fc7478f981a28880f2e4e07,

title = "Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task",

abstract = "Robustness of different pattern recognition methods is one of the key challenges in autonomous driving, especially when driving in the high variety of road environments and weather conditions, such as gravel roads and snowfall. Although one can collect data from these adverse conditions using cars equipped with sensors, it is quite tedious to annotate the data for training. In this work, we address this limitation and propose a CNN-based method that can leverage the steering wheel angle information to improve the road area semantic segmentation. As the steering wheel angle data can be easily acquired with the associated images, one could improve the accuracy of road area semantic segmentation by collecting data in new road environments without manual data annotation. We demonstrate the effectiveness of the proposed approach on two challenging data sets for autonomous driving and show that when the steering task is used in our segmentation model training, it leads to a 0.1–2.9% gain in the road area mIoU (mean Intersection over Union) compared to the corresponding reference transfer learning model.",

keywords = "Autonomous driving, Domain adaptation, Multi-task learning, Road area semantic segmentation, Transfer learning",

author = "Jyri Maanp{\"a}{\"a} and Iaroslav Melekhov and Josef Taher and Petri Manninen and Juha Hyypp{\"a}",

note = "Funding Information: Acknowledgement. Academy of Finland projects (decisions 318437 and 319011) and Henry Ford foundation are gratefully acknowledged for financial support. Authors{\textquoteright} contribution is the following: Maanp{\"a}{\"a} designed and performed the experiments and wrote the manuscript. Melekhov advised in the model development and provided feedback on the manuscript. Maanp{\"a}{\"a}, Taher, and Manninen took equal shares in instrument- ing the autonomous driving platform and developing software. Maanp{\"a}{\"a} and Taher participated in data collection. Hyypp{\"a} supervised the project. Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; International Conference on Image Analysis and Processing, ICIAP ; Conference date: 23-05-2022 Through 27-05-2022",

year = "2022",

doi = "10.1007/978-3-031-06427-2_60",

language = "English",

isbn = "978-3-031-06426-5",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer",

pages = "727--738",

editor = "Stan Sclaroff and Cosimo Distante and Marco Leo and Farinella, {Giovanni M.} and Federico Tombari",

booktitle = "Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings",

address = "Germany",

}

Maanpää, J , Melekhov, I , Taher, J , Manninen, P & Hyyppä, J 2022, Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task. in S Sclaroff, C Distante, M Leo, GM Farinella & F Tombari (eds), Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13231 LNCS, Springer, pp. 727-738, International Conference on Image Analysis and Processing, Lecce, Italy, 23/05/2022. https://doi.org/10.1007/978-3-031-06427-2_60

Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task. / Maanpää, Jyri ; Melekhov, Iaroslav ; Taher, Josef et al.
Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings. ed. / Stan Sclaroff; Cosimo Distante; Marco Leo; Giovanni M. Farinella; Federico Tombari. Springer, 2022. p. 727-738 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13231 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task

AU - Maanpää, Jyri

AU - Melekhov, Iaroslav

AU - Taher, Josef

AU - Manninen, Petri

AU - Hyyppä, Juha

N1 - Conference code: 21

PY - 2022

Y1 - 2022

N2 - Robustness of different pattern recognition methods is one of the key challenges in autonomous driving, especially when driving in the high variety of road environments and weather conditions, such as gravel roads and snowfall. Although one can collect data from these adverse conditions using cars equipped with sensors, it is quite tedious to annotate the data for training. In this work, we address this limitation and propose a CNN-based method that can leverage the steering wheel angle information to improve the road area semantic segmentation. As the steering wheel angle data can be easily acquired with the associated images, one could improve the accuracy of road area semantic segmentation by collecting data in new road environments without manual data annotation. We demonstrate the effectiveness of the proposed approach on two challenging data sets for autonomous driving and show that when the steering task is used in our segmentation model training, it leads to a 0.1–2.9% gain in the road area mIoU (mean Intersection over Union) compared to the corresponding reference transfer learning model.

AB - Robustness of different pattern recognition methods is one of the key challenges in autonomous driving, especially when driving in the high variety of road environments and weather conditions, such as gravel roads and snowfall. Although one can collect data from these adverse conditions using cars equipped with sensors, it is quite tedious to annotate the data for training. In this work, we address this limitation and propose a CNN-based method that can leverage the steering wheel angle information to improve the road area semantic segmentation. As the steering wheel angle data can be easily acquired with the associated images, one could improve the accuracy of road area semantic segmentation by collecting data in new road environments without manual data annotation. We demonstrate the effectiveness of the proposed approach on two challenging data sets for autonomous driving and show that when the steering task is used in our segmentation model training, it leads to a 0.1–2.9% gain in the road area mIoU (mean Intersection over Union) compared to the corresponding reference transfer learning model.

KW - Autonomous driving

KW - Domain adaptation

KW - Multi-task learning

KW - Road area semantic segmentation

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85130888856&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-06427-2_60

DO - 10.1007/978-3-031-06427-2_60

M3 - Conference article in proceedings

AN - SCOPUS:85130888856

SN - 978-3-031-06426-5

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 727

EP - 738

BT - Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings

A2 - Sclaroff, Stan

A2 - Distante, Cosimo

A2 - Leo, Marco

A2 - Farinella, Giovanni M.

A2 - Tombari, Federico

PB - Springer

T2 - International Conference on Image Analysis and Processing

Y2 - 23 May 2022 through 27 May 2022

ER -

Maanpää J , Melekhov I , Taher J , Manninen P , Hyyppä J. Leveraging Road Area Semantic Segmentation with Auxiliary Steering Task. In Sclaroff S, Distante C, Leo M, Farinella GM, Tombari F, editors, Image Analysis and Processing – ICIAP 2022 - 21st International Conference, 2022, Proceedings. Springer. 2022. p. 727-738. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-06427-2_60