SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation

Ysobel Sims; Trent Houliston; Matthew Amos; Alexander Biddulph; Jonathan Tabac; Alana Noonan; Angelique Herfel; Joe Bailey; Johanne Montano; Thomas O’Brien

doi:10.1007/978-3-031-85859-8_27

SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation

Ysobel Sims^*, Trent Houliston, Matthew Amos, Alexander Biddulph, Jonathan Tabac, Alana Noonan, Angelique Herfel, Joe Bailey, Johanne Montano, Thomas O’Brien

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding contribution › peer-review

Abstract

SoS (Soccer Segmentation) is a computer vision semantic segmentation dataset for RoboCup soccer. It contains raw images with rectilinear and equisolid projections, metadata for each image with lens and positional information, and coloured segmentation masks. The dataset is generated using Blender and uses 360^∘ images from real past RoboCup fields to create a semi-synthetic scene. Our Blender tool used to generate the dataset is available at https://github.com/NUbots/NUpbr. SoS is hosted on HuggingFace and is accessible for download at https://doi.org/10.57967/hf/2099. We provide benchmarks for SoS using the Visual Mesh and U-Net, and conduct real world experiments on the network with fine-tuning on a small amount of real data.

Original language	English
Title of host publication	RoboCup 2024
Subtitle of host publication	Robot World Cup XXVII
Editors	Edna Barros, Josiah P. Hanna, Hiroyuki Okada, Elena Torta
Place of Publication	Cham
Publisher	Springer, Springer Nature
Pages	318-327
Number of pages	10
ISBN (Electronic)	9783031858598
ISBN (Print)	9783031858581
DOIs	https://doi.org/10.1007/978-3-031-85859-8_27
Publication status	Published - 2025
Externally published	Yes
Event	27th RoboCup International Symposium, 2024 - Eindhoven, Netherlands Duration: 15 Jul 2024 → 22 Jul 2024

Publication series

Name	Lecture Notes in Computer Science
Volume	15570 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	27th RoboCup International Symposium, 2024
Country/Territory	Netherlands
City	Eindhoven
Period	15/07/24 → 22/07/24

Keywords

Computer Vision
Dataset
Machine Learning

Access to Document

10.1007/978-3-031-85859-8_27

Cite this

Sims, Y., Houliston, T., Amos, M., Biddulph, A., Tabac, J., Noonan, A., Herfel, A., Bailey, J., Montano, J., & O’Brien, T. (2025). SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation. In E. Barros, J. P. Hanna, H. Okada, & E. Torta (Eds.), RoboCup 2024: Robot World Cup XXVII (pp. 318-327). (Lecture Notes in Computer Science; Vol. 15570 LNAI). Springer, Springer Nature. https://doi.org/10.1007/978-3-031-85859-8_27

@inproceedings{7925c0fe692d47eba71682117c45a13d,

title = "SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation",

abstract = "SoS (Soccer Segmentation) is a computer vision semantic segmentation dataset for RoboCup soccer. It contains raw images with rectilinear and equisolid projections, metadata for each image with lens and positional information, and coloured segmentation masks. The dataset is generated using Blender and uses 360∘ images from real past RoboCup fields to create a semi-synthetic scene. Our Blender tool used to generate the dataset is available at https://github.com/NUbots/NUpbr. SoS is hosted on HuggingFace and is accessible for download at https://doi.org/10.57967/hf/2099. We provide benchmarks for SoS using the Visual Mesh and U-Net, and conduct real world experiments on the network with fine-tuning on a small amount of real data.",

keywords = "Computer Vision, Dataset, Machine Learning",

author = "Ysobel Sims and Trent Houliston and Matthew Amos and Alexander Biddulph and Jonathan Tabac and Alana Noonan and Angelique Herfel and Joe Bailey and Johanne Montano and Thomas O{\textquoteright}Brien",

year = "2025",

doi = "10.1007/978-3-031-85859-8\_27",

language = "English",

isbn = "9783031858581",

series = "Lecture Notes in Computer Science",

publisher = "Springer, Springer Nature",

pages = "318--327",

editor = "Edna Barros and Hanna, \{Josiah P.\} and Hiroyuki Okada and Elena Torta",

booktitle = "RoboCup 2024",

address = "United States",

note = "27th RoboCup International Symposium, 2024 ; Conference date: 15-07-2024 Through 22-07-2024",

}

Sims, Y, Houliston, T, Amos, M, Biddulph, A, Tabac, J, Noonan, A, Herfel, A, Bailey, J, Montano, J & O’Brien, T 2025, SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation. in E Barros, JP Hanna, H Okada & E Torta (eds), RoboCup 2024: Robot World Cup XXVII. Lecture Notes in Computer Science, vol. 15570 LNAI, Springer, Springer Nature, Cham, pp. 318-327, 27th RoboCup International Symposium, 2024, Eindhoven, Netherlands, 15/07/24. https://doi.org/10.1007/978-3-031-85859-8_27

SoS: a semi-synthetic RoboCup soccer dataset for visual segmentation. / Sims, Ysobel; Houliston, Trent; Amos, Matthew et al.
RoboCup 2024: Robot World Cup XXVII. ed. / Edna Barros; Josiah P. Hanna; Hiroyuki Okada; Elena Torta. Cham: Springer, Springer Nature, 2025. p. 318-327 (Lecture Notes in Computer Science; Vol. 15570 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference proceeding contribution › peer-review

TY - GEN

T1 - SoS

T2 - 27th RoboCup International Symposium, 2024

AU - Sims, Ysobel

AU - Houliston, Trent

AU - Amos, Matthew

AU - Biddulph, Alexander

AU - Tabac, Jonathan

AU - Noonan, Alana

AU - Herfel, Angelique

AU - Bailey, Joe

AU - Montano, Johanne

AU - O’Brien, Thomas

PY - 2025

Y1 - 2025

N2 - SoS (Soccer Segmentation) is a computer vision semantic segmentation dataset for RoboCup soccer. It contains raw images with rectilinear and equisolid projections, metadata for each image with lens and positional information, and coloured segmentation masks. The dataset is generated using Blender and uses 360∘ images from real past RoboCup fields to create a semi-synthetic scene. Our Blender tool used to generate the dataset is available at https://github.com/NUbots/NUpbr. SoS is hosted on HuggingFace and is accessible for download at https://doi.org/10.57967/hf/2099. We provide benchmarks for SoS using the Visual Mesh and U-Net, and conduct real world experiments on the network with fine-tuning on a small amount of real data.

AB - SoS (Soccer Segmentation) is a computer vision semantic segmentation dataset for RoboCup soccer. It contains raw images with rectilinear and equisolid projections, metadata for each image with lens and positional information, and coloured segmentation masks. The dataset is generated using Blender and uses 360∘ images from real past RoboCup fields to create a semi-synthetic scene. Our Blender tool used to generate the dataset is available at https://github.com/NUbots/NUpbr. SoS is hosted on HuggingFace and is accessible for download at https://doi.org/10.57967/hf/2099. We provide benchmarks for SoS using the Visual Mesh and U-Net, and conduct real world experiments on the network with fine-tuning on a small amount of real data.

KW - Computer Vision

KW - Dataset

KW - Machine Learning

UR - https://www.scopus.com/pages/publications/105003858801

U2 - 10.1007/978-3-031-85859-8_27

DO - 10.1007/978-3-031-85859-8_27

M3 - Conference proceeding contribution

AN - SCOPUS:105003858801

SN - 9783031858581

T3 - Lecture Notes in Computer Science

SP - 318

EP - 327

BT - RoboCup 2024

A2 - Barros, Edna

A2 - Hanna, Josiah P.

A2 - Okada, Hiroyuki

A2 - Torta, Elena

PB - Springer, Springer Nature

CY - Cham

Y2 - 15 July 2024 through 22 July 2024

ER -