Learning Regional Attraction for Line Segment Detection

Xue, Nan; Bai, Song; Wang, Fu-Dong; Xia, Gui-Song; Wu, Tianfu; Zhang, Liangpei; Torr, Philip H. S.

doi:10.1109/tpami.2019.2958642

Cited by 30 publications

(13 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Detection of connected line segments has been studied as the task of Wireframe Estimation [2], [38]- [40]. However, these methods do not try to use semantic understanding of the lines and junctions and do not exploit such semantic understanding in terms of Room Layout Estimation.…”

Section: Related Workmentioning

confidence: 99%

Semantic Room Wireframe Detection from a Single View

Gillsjö¹,

Flood²,

Åström³

2022

Preprint

View full text Add to dashboard Cite

Fig. 1. The images on top show the desired results from Semantic Room Wireframe detection and below are the results from a Room Layout estimation algorithm [1]. Lines are added between all detected planes for comparison.The algorithm produces satisfying results on simple scenes (left) but is lacking on more complex scenes, e.g. to the right where the wide opening is erroneously classified as part of the wall.

show abstract

Section: Related Workmentioning

confidence: 99%

Semantic Room Wireframe Detection from a Single View

Gillsjö¹,

Flood²,

Åström³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Benefiting from the line-rich characteristics of our proposed SLF dataset, we are able to measure the geometric accuracy by comparing the rectified heatmaps of distorted lines with the groundtruth of line segments. Motivated by the evaluation protocols used for edge detection [46] and line segment detection [29], [30], we use the precision and recall to measure if the pixels on the distorted lines are still on the straight lines after rectification. Denoting the rectified heatmap of the distorted lines by Lr , we use the matching strategy proposed in [46] to match the edge pixels with the (e) Rong [24] (f) Ours (g) GT Fig.…”

Section: Precision-vs-recall For the Rectified Heatmaps Of Linesmentioning

confidence: 99%

Fisheye Distortion Rectification from Deep Straight Lines

Xue¹,

Xue²,

Xia³

2020

Preprint

Self Cite

View full text Add to dashboard Cite

This paper presents a novel line-aware rectification network (LaRecNet) to address the problem of fisheye distortion rectification based on the classical observation that straight lines in 3D space should be still straight in image planes. Specifically, the proposed LaRecNet contains three sequential modules to (1) learn the distorted straight lines from fisheye images; (2) estimate the distortion parameters from the learned heatmaps and the image appearance; and (3) rectify the input images via a proposed differentiable rectification layer. To better train and evaluate the proposed model, we create a synthetic line-rich fisheye (SLF) dataset that contains the distortion parameters and well-annotated distorted straight lines of fisheye images. The proposed method enables us to simultaneously calibrate the geometric distortion parameters and rectify fisheye images. Extensive experiments demonstrate that our model achieves state-of-the-art performance in terms of both geometric accuracy and image quality on several evaluation metrics. In particular, the images rectified by LaRecNet achieve an average reprojection error of 0.33 pixels on the SLF dataset and produce the highest peak signal-to-noise ratio (PSNR) and structure similarity index (SSIM) compared with the groundtruth.

show abstract

“…Although there are various low-level geometric primitives, we find that line segments are usually used to construct 3D planes [16,25] and contain more holistic 3D information of the scene when comparing with other geometric primitives, such as feature points, edges, and vanishing points. Besides, benefiting from recent state-of-the-art works in line segment detection [42,40,41,47], it is convenient for us to achieve line segments from an image. Thus, in this paper, we use line segments as the geometric structures for plane recovery.…”

Section: Introductionmentioning

confidence: 99%

PlaneTR: Structure-Guided Transformers for 3D Plane Recovery

Tan¹,

Xue

Bai

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

This paper presents a neural network built upon Transformers, namely PlaneTR, to simultaneously detect and reconstruct planes from a single image. Different from previous methods, PlaneTR jointly leverages the context information and the geometric structures in a sequence-tosequence way to holistically detect plane instances in one forward pass. Specifically, we represent the geometric structures as line segments and conduct the network with three main components: (i) context and line segments encoders, (ii) a structure-guided plane decoder, (iii) a pixelwise plane embedding decoder. Given an image and its detected line segments, PlaneTR generates the context and line segment sequences via two specially designed encoders and then feeds them into a Transformers-based decoder to directly predict a sequence of plane instances by simultaneously considering the context and global structure cues. Finally, the pixel-wise embeddings are computed to assign each pixel to one predicted plane instance which is nearest to it in embedding space. Comprehensive experiments demonstrate that PlaneTR achieves a state-of-the-art performance on the ScanNet and NYUv2 datasets.

show abstract

Learning Regional Attraction for Line Segment Detection

Cited by 30 publications

References 52 publications

Semantic Room Wireframe Detection from a Single View

Semantic Room Wireframe Detection from a Single View

Fisheye Distortion Rectification from Deep Straight Lines

PlaneTR: Structure-Guided Transformers for 3D Plane Recovery

Contact Info

Product

Resources

About