Semantic Room Wireframe Detection from a Single View

David Gillsjö; Gabrielle Flood; Kalle Åström

doi:10.1109/ICPR56361.2022.9956252

Semantic Room Wireframe Detection from a Single View

David Gillsjö, Gabrielle Flood, Kalle Åström

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

Abstract

Reconstruction of indoor surfaces with limited texture information or with repeated textures, a situation common in walls and ceilings, may be difficult with a monocular Structure from Motion system. We propose a Semantic Room Wireframe Detection task to predict a Semantic Wireframe from a single perspective image. Such predictions may be used with shape priors to estimate the Room Layout and aid reconstruction. To train and test the proposed algorithm we create a new set of annotations from the simulated Structured3D dataset. We show qualitatively that the SRW-Net handles complex room geometries better than previous Room Layout Estimation algorithms while quantitatively out-performing the baseline in non-semantic Wireframe Detection.

Original language	English
Title of host publication	26th International Conference on Pattern Recognition, 2022
Publisher	IEEE - Institute of Electrical and Electronics Engineers Inc.
Pages	1886-1893
ISBN (Electronic)	978-1-6654-9062-7
ISBN (Print)	978-1-6654-9063-4
DOIs	https://doi.org/10.1109/ICPR56361.2022.9956252
Publication status	Published - 2022 Aug 25
Event	26TH International Conference on Pattern Recognition, 2022 - Montreal, Canada Duration: 2022 Aug 21 → 2022 Aug 25

Conference

Conference	26TH International Conference on Pattern Recognition, 2022
Abbreviated title	ICPR 2022
Country/Territory	Canada
City	Montreal
Period	2022/08/21 → 2022/08/25

Subject classification (UKÄ)

Computer Vision and Robotics (Autonomous Systems)

Access to Document

10.1109/ICPR56361.2022.9956252

https://arxiv.org/abs/2206.00491Licence: Other

1 Doctoral Thesis (compilation)

Applications in Monocular Computer Vision using Geometry and Learning: Map Merging, 3D Reconstruction and Detection of Geometric Primitives
Gillsjö, D., 2023, Lund: Lund University / Centre for Mathematical Sciences /LTH.
Research output: Thesis › Doctoral Thesis (compilation)

Open Access
File

1 Active

Semantic Structure from Motion
Gillsjö, D., Åström, K. & Heyden, A.
2016/08/01 → …
Project: Dissertation

Cite this

@inproceedings{9852dc4ec1524a048ecddc8cf4b751f3,

title = "Semantic Room Wireframe Detection from a Single View",

abstract = "Reconstruction of indoor surfaces with limited texture information or with repeated textures, a situation common in walls and ceilings, may be difficult with a monocular Structure from Motion system. We propose a Semantic Room Wireframe Detection task to predict a Semantic Wireframe from a single perspective image. Such predictions may be used with shape priors to estimate the Room Layout and aid reconstruction. To train and test the proposed algorithm we create a new set of annotations from the simulated Structured3D dataset. We show qualitatively that the SRW-Net handles complex room geometries better than previous Room Layout Estimation algorithms while quantitatively out-performing the baseline in non-semantic Wireframe Detection.",

author = "David Gillsj{\"o} and Gabrielle Flood and Kalle {\AA}str{\"o}m",

year = "2022",

month = aug,

day = "25",

doi = "10.1109/ICPR56361.2022.9956252",

language = "English",

isbn = "978-1-6654-9063-4",

pages = "1886--1893",

booktitle = "26th International Conference on Pattern Recognition, 2022",

publisher = "IEEE - Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "26TH International Conference on Pattern Recognition, 2022 ; Conference date: 21-08-2022 Through 25-08-2022",

}

Gillsjö, D , Flood, G & Åström, K 2022, Semantic Room Wireframe Detection from a Single View. in 26th International Conference on Pattern Recognition, 2022. IEEE - Institute of Electrical and Electronics Engineers Inc., pp. 1886-1893, 26TH International Conference on Pattern Recognition, 2022, Montreal, Quebec, Canada, 2022/08/21. https://doi.org/10.1109/ICPR56361.2022.9956252

TY - GEN

T1 - Semantic Room Wireframe Detection from a Single View

AU - Gillsjö, David

AU - Flood, Gabrielle

AU - Åström, Kalle

PY - 2022/8/25

Y1 - 2022/8/25

N2 - Reconstruction of indoor surfaces with limited texture information or with repeated textures, a situation common in walls and ceilings, may be difficult with a monocular Structure from Motion system. We propose a Semantic Room Wireframe Detection task to predict a Semantic Wireframe from a single perspective image. Such predictions may be used with shape priors to estimate the Room Layout and aid reconstruction. To train and test the proposed algorithm we create a new set of annotations from the simulated Structured3D dataset. We show qualitatively that the SRW-Net handles complex room geometries better than previous Room Layout Estimation algorithms while quantitatively out-performing the baseline in non-semantic Wireframe Detection.

AB - Reconstruction of indoor surfaces with limited texture information or with repeated textures, a situation common in walls and ceilings, may be difficult with a monocular Structure from Motion system. We propose a Semantic Room Wireframe Detection task to predict a Semantic Wireframe from a single perspective image. Such predictions may be used with shape priors to estimate the Room Layout and aid reconstruction. To train and test the proposed algorithm we create a new set of annotations from the simulated Structured3D dataset. We show qualitatively that the SRW-Net handles complex room geometries better than previous Room Layout Estimation algorithms while quantitatively out-performing the baseline in non-semantic Wireframe Detection.

U2 - 10.1109/ICPR56361.2022.9956252

DO - 10.1109/ICPR56361.2022.9956252

M3 - Paper in conference proceeding

SN - 978-1-6654-9063-4

SP - 1886

EP - 1893

BT - 26th International Conference on Pattern Recognition, 2022

PB - IEEE - Institute of Electrical and Electronics Engineers Inc.

T2 - 26TH International Conference on Pattern Recognition, 2022

Y2 - 21 August 2022 through 25 August 2022

ER -

Semantic Room Wireframe Detection from a Single View

Abstract

Conference

Subject classification (UKÄ)

Access to Document

Fingerprint

Research output

Applications in Monocular Computer Vision using Geometry and Learning: Map Merging, 3D Reconstruction and Detection of Geometric Primitives

Projects

Semantic Structure from Motion

Cite this