CurveNet: Curvature-Based Multitask Learning Deep Networks for 3D Object Recognition

Muzahid, Abdullah; Wan, Wanggen; Sohel, Ferdous; Wu, Lianyao; Hou, Long

doi:10.1109/jas.2020.1003324

Cited by 54 publications

(23 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pointnet [12] xyz 1024 89.2 Pointnet++ [13] xyz 1024 90.7 Kd-Net [21] xyz 32k 91.8 DGCNN [40] xyz 1024 92.9 SRN [14] xyz 1024 91.5 PointGrid [11] xyz 1024 92.0 PointCNN [41] xyz 1024 92.2 RS-CNN [42] xyz 1024 93.6 PCT [33] xyz 1024 93.6 PAConv [43] xyz 1024 93.9 CurveNet [44] xyz 1024 93.8 RPNet-W9 [45] xyz 1024 93.9 Pointnet++ [13] xyz,nr 1024 91.7 PAT [16] xyz,nr 1024 91.7 SpiderCNN [46] xyz,nr 5k 92.4 A-CNN [47] xyz,nr 1024 92.6 PointASNL [48] xyz,nr 1024 93.2 SO-Net [49] xyz,nr 1024 93.4 PointSCNet xyz,nr 1024 93.7…”

Section: Methods Input Points Accmentioning

confidence: 99%

PointSCNet: Point Cloud Structure and Correlation Learning Based on Space-Filling Curve-Guided Sampling

Chen

et al. 2021

Symmetry

View full text Add to dashboard Cite

Geometrical structures and the internal local region relationship, such as symmetry, regular array, junction, etc., are essential for understanding a 3D shape. This paper proposes a point cloud feature extraction network named PointSCNet, to capture the geometrical structure information and local region correlation information of a point cloud. The PointSCNet consists of three main modules: the space-filling curve-guided sampling module, the information fusion module, and the channel-spatial attention module. The space-filling curve-guided sampling module uses Z-order curve coding to sample points that contain geometrical correlation. The information fusion module uses a correlation tensor and a set of skip connections to fuse the structure and correlation information. The channel-spatial attention module enhances the representation of key points and crucial feature channels to refine the network. The proposed PointSCNet is evaluated on shape classification and part segmentation tasks. The experimental results demonstrate that the PointSCNet outperforms or is on par with state-of-the-art methods by learning the structure and correlation of point clouds effectively.

show abstract

Section: Methods Input Points Accmentioning

confidence: 99%

PointSCNet: Point Cloud Structure and Correlation Learning Based on Space-Filling Curve-Guided Sampling

Chen

et al. 2021

Symmetry

View full text Add to dashboard Cite

show abstract

“…The more general field of computer vision has experienced a fundamental transition towards data-hungry deep learning methods thanks to their natural ability to process data in raw form [32]. The transition started with 2D tasks, such as object detection [33], [34] and human pose estimation [35], and it expanded to 3D tasks such as 3D object detection [36]- [38], object recognition [39], depth estimation [40], or even forecasting tasks [41]. A crucial factor in this transformation has been the release of massive datasets for 2D [42]- [44] and 3D tasks [30], [45]- [48], especially in the context of autonomous driving.…”

Section: Related Workmentioning

confidence: 99%

Perceiving Humans: From Monocular 3D Localization to Social Distancing

Bertoni

Kreiss

Alahi

2022

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

Perceiving humans in the context of Intelligent Transportation Systems (ITS) often relies on multiple cameras or expensive LiDAR sensors. In this work, we present a new cost-effective vision-based method that perceives humans' locations in 3D and their body orientation from a single image. We address the challenges related to the ill-posed monocular 3D tasks by proposing a neural network architecture that predicts confidence intervals in contrast to point estimates. Our neural network estimates human 3D body locations and their orientation with a measure of uncertainty. Our proposed solution (i) is privacy-safe, (ii) works with any fixed or moving cameras, and (iii) does not rely on ground plane estimation. We demonstrate the performance of our method with respect to three applications: locating humans in 3D, detecting social interactions, and verifying the compliance of recent safety measures due to the COVID-19 outbreak. We show that it is possible to rethink the concept of "social distancing" as a form of social interaction in contrast to a simple location-based rule. We publicly share the source code towards an open science mission.

show abstract

“…For instance, this kind of defect inspection is performed using near-netshape production techniques [16] and the kriging model with statistical models to compute the shape deviation errors [17,18]. The second category uses deep computer vision models trained on a labeled dataset, taking advantage of the rapid progress of this field [19][20][21]. For instance, some approaches use local binary patterns [22], Class-balanced Hierarchical Refinement (CHR) [23], Convolutional Neural Networks (CNNs) [24], and spatial attention bilinear CNNs [25].…”

Section: Introductionmentioning

confidence: 99%

CTIMS: Automated Defect Detection Framework Using Computed Tomography

et al. 2022

View full text Add to dashboard Cite

Non-Destructive Testing (NDT) is one of the inspection techniques used in industrial tool inspection for quality and safety control. It is performed mainly using X-ray Computed Tomography (CT) to scan the internal structure of the tools and detect the potential defects. In this paper, we propose a new toolbox called the CT-Based Integrity Monitoring System (CTIMS-Toolbox) for automated inspection of CT images and volumes. It contains three main modules: first, the database management module, which handles the database and reads/writes queries to retrieve or save the CT data; second, the pre-processing module for registration and background subtraction; third, the defect inspection module to detect all the potential defects (missing parts, damaged screws, etc.) based on a hybrid system composed of computer vision and deep learning techniques. This paper explores the different features of the CTIMS-Toolbox, exposes the performance of its modules, compares its features to some existing CT inspection toolboxes, and provides some examples of the obtained results.

show abstract

CurveNet: Curvature-Based Multitask Learning Deep Networks for 3D Object Recognition

Cited by 54 publications

References 40 publications

PointSCNet: Point Cloud Structure and Correlation Learning Based on Space-Filling Curve-Guided Sampling

PointSCNet: Point Cloud Structure and Correlation Learning Based on Space-Filling Curve-Guided Sampling

Perceiving Humans: From Monocular 3D Localization to Social Distancing

CTIMS: Automated Defect Detection Framework Using Computed Tomography

Contact Info

Product

Resources

About