BS3D: Building-Scale 3D Reconstruction from RGB-D Images

Janne Mustaniemi; Juho Kannala; Esa Rahtu; Li Liu; Janne Heikkilä

doi:10.1007/978-3-031-31438-4_36

BS3D: Building-Scale 3D Reconstruction from RGB-D Images

Janne Mustaniemi^*
, Juho Kannala
, Esa Rahtu
, Li Liu
, Janne Heikkilä

^*Corresponding author for this work

Computing Sciences

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

3 Citations (Scopus)

Abstract

Various datasets have been proposed for simultaneous localization and mapping (SLAM) and related problems. Existing datasets often include small environments, have incomplete ground truth, or lack important sensor data, such as depth and infrared images. We propose an easy-to-use framework for acquiring building-scale 3D reconstruction using a consumer depth camera. Unlike complex and expensive acquisition setups, our system enables crowd-sourcing, which can greatly benefit data-hungry algorithms. Compared to similar systems, we utilize raw depth maps for odometry computation and loop closure refinement which results in better reconstructions. We acquire a building-scale 3D dataset (BS3D) and demonstrate its value by training an improved monocular depth estimation model. As a unique experiment, we benchmark visual-inertial odometry methods using both color and active infrared images.

Original language	English
Title of host publication	Image Analysis
Subtitle of host publication	22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II
Editors	Rikke Gade, Michael Felsberg, Joni-Kristian Kämäräinen
Publisher	Springer
Pages	551-565
Number of pages	15
ISBN (Print)	9783031314377
DOIs	https://doi.org/10.1007/978-3-031-31438-4_36
Publication status	Published - 2023
Publication type	A4 Article in conference proceedings
Event	Scandinavian Conference on Image Analysis - Lapland, Finland Duration: 18 Apr 2023 → 21 Apr 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13886 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Scandinavian Conference on Image Analysis
Country/Territory	Finland
City	Lapland
Period	18/04/23 → 21/04/23

Keywords

Depth camera
Large-scale
SLAM

Publication forum classification

Publication forum level 1

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-31438-4_36

https://urn.fi/URN:NBN:fi-fe2023062057123

Cite this

Mustaniemi, J., Kannala, J., Rahtu, E., Liu, L., & Heikkilä, J. (2023). BS3D: Building-Scale 3D Reconstruction from RGB-D Images. In R. Gade, M. Felsberg, & J.-K. Kämäräinen (Eds.), Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II (pp. 551-565). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13886 LNCS). Springer. https://doi.org/10.1007/978-3-031-31438-4_36

Mustaniemi, Janne ; Kannala, Juho ; Rahtu, Esa et al. / BS3D : Building-Scale 3D Reconstruction from RGB-D Images. Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II. editor / Rikke Gade ; Michael Felsberg ; Joni-Kristian Kämäräinen. Springer, 2023. pp. 551-565 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{09f0746177fb43a4a114f25b928def89,

title = "BS3D: Building-Scale 3D Reconstruction from RGB-D Images",

abstract = "Various datasets have been proposed for simultaneous localization and mapping (SLAM) and related problems. Existing datasets often include small environments, have incomplete ground truth, or lack important sensor data, such as depth and infrared images. We propose an easy-to-use framework for acquiring building-scale 3D reconstruction using a consumer depth camera. Unlike complex and expensive acquisition setups, our system enables crowd-sourcing, which can greatly benefit data-hungry algorithms. Compared to similar systems, we utilize raw depth maps for odometry computation and loop closure refinement which results in better reconstructions. We acquire a building-scale 3D dataset (BS3D) and demonstrate its value by training an improved monocular depth estimation model. As a unique experiment, we benchmark visual-inertial odometry methods using both color and active infrared images.",

keywords = "Depth camera, Large-scale, SLAM",

author = "Janne Mustaniemi and Juho Kannala and Esa Rahtu and Li Liu and Janne Heikkil{\"a}",

note = "Publisher Copyright: {\textcopyright} 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.; Scandinavian Conference on Image Analysis ; Conference date: 18-04-2023 Through 21-04-2023",

year = "2023",

doi = "10.1007/978-3-031-31438-4\_36",

language = "English",

isbn = "9783031314377",

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

publisher = "Springer",

pages = "551--565",

editor = "Rikke Gade and Michael Felsberg and Joni-Kristian K{\"a}m{\"a}r{\"a}inen",

booktitle = "Image Analysis",

}

Mustaniemi, J, Kannala, J, Rahtu, E, Liu, L & Heikkilä, J 2023, BS3D: Building-Scale 3D Reconstruction from RGB-D Images. in R Gade, M Felsberg & J-K Kämäräinen (eds), Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13886 LNCS, Springer, pp. 551-565, Scandinavian Conference on Image Analysis, Lapland, Finland, 18/04/23. https://doi.org/10.1007/978-3-031-31438-4_36

BS3D: Building-Scale 3D Reconstruction from RGB-D Images. / Mustaniemi, Janne; Kannala, Juho; Rahtu, Esa et al.
Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II. ed. / Rikke Gade; Michael Felsberg; Joni-Kristian Kämäräinen. Springer, 2023. p. 551-565 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13886 LNCS).

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

TY - GEN

T1 - BS3D

T2 - Scandinavian Conference on Image Analysis

AU - Mustaniemi, Janne

AU - Kannala, Juho

AU - Rahtu, Esa

AU - Liu, Li

AU - Heikkilä, Janne

PY - 2023

Y1 - 2023

N2 - Various datasets have been proposed for simultaneous localization and mapping (SLAM) and related problems. Existing datasets often include small environments, have incomplete ground truth, or lack important sensor data, such as depth and infrared images. We propose an easy-to-use framework for acquiring building-scale 3D reconstruction using a consumer depth camera. Unlike complex and expensive acquisition setups, our system enables crowd-sourcing, which can greatly benefit data-hungry algorithms. Compared to similar systems, we utilize raw depth maps for odometry computation and loop closure refinement which results in better reconstructions. We acquire a building-scale 3D dataset (BS3D) and demonstrate its value by training an improved monocular depth estimation model. As a unique experiment, we benchmark visual-inertial odometry methods using both color and active infrared images.

AB - Various datasets have been proposed for simultaneous localization and mapping (SLAM) and related problems. Existing datasets often include small environments, have incomplete ground truth, or lack important sensor data, such as depth and infrared images. We propose an easy-to-use framework for acquiring building-scale 3D reconstruction using a consumer depth camera. Unlike complex and expensive acquisition setups, our system enables crowd-sourcing, which can greatly benefit data-hungry algorithms. Compared to similar systems, we utilize raw depth maps for odometry computation and loop closure refinement which results in better reconstructions. We acquire a building-scale 3D dataset (BS3D) and demonstrate its value by training an improved monocular depth estimation model. As a unique experiment, we benchmark visual-inertial odometry methods using both color and active infrared images.

KW - Depth camera

KW - Large-scale

KW - SLAM

U2 - 10.1007/978-3-031-31438-4_36

DO - 10.1007/978-3-031-31438-4_36

M3 - Conference contribution

AN - SCOPUS:85161447831

SN - 9783031314377

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

SP - 551

EP - 565

BT - Image Analysis

A2 - Gade, Rikke

A2 - Felsberg, Michael

A2 - Kämäräinen, Joni-Kristian

PB - Springer

Y2 - 18 April 2023 through 21 April 2023

ER -

Mustaniemi J, Kannala J, Rahtu E, Liu L, Heikkilä J. BS3D: Building-Scale 3D Reconstruction from RGB-D Images. In Gade R, Felsberg M, Kämäräinen JK, editors, Image Analysis: 22nd Scandinavian Conference, SCIA 2023, Sirkka, Finland, April 18–21, 2023, Proceedings, Part II. Springer. 2023. p. 551-565. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-31438-4_36