Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data

Uy, Mikaela Angelina; Pham, Quang-Hieu; Hua, Binh-Son; Nguyen, Thanh; Yeung, Sai-Kit

doi:10.1109/iccv.2019.00167

Cited by 491 publications

(361 citation statements)

References 41 publications

Supporting

Mentioning

319

Contrasting

Unclassified

Order By: Relevance

“…Second, generalizing point cloud convolutions and object classification to support non-rigid transformations and deformable objects could further improve overall robustness. Finally, more thorough benchmarking rotation-invariant convolutions with real-world data [36] is necessary to understand the impact of such data on the learning of rotation-invariant features.…”

Section: Resultsmentioning

confidence: 99%

“…Recently, there have been significant advances in applying deep learning [19] to train neural networks for numerous tasks such as object classification and semantic segmentation. With the wide availability of consumer-grade depth sensors, acquiring 3D data has become more intuitive and robust with many 3D datasets available publicly [40,4,14,7,2,43,36]. This leads to increased interests in tackling scene understanding in the 3D Among the representations for 3D data, a promising direction is to let neural networks consume point cloud data directly since point cloud data is the common data format acquired from depth sensors such as RGB-D or LiDAR cameras.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rotation Invariant Convolutions for 3D Point Clouds Deep Learning

Zhang

Hua

Rosen

et al. 2019

2019 International Conference on 3D Vision (3DV)

Self Cite

136

144

View full text Add to dashboard Cite

Recent progresses in 3D deep learning has shown that it is possible to design special convolution operators to consume point cloud data. However, a typical drawback is that rotation invariance is often not guaranteed, resulting in networks that generalizes poorly to arbitrary rotations. In this paper, we introduce a novel convolution operator for point clouds that achieves rotation invariance. Our core idea is to use lowlevel rotation invariant geometric features such as distances and angles to design a convolution operator for point cloud learning. The well-known point ordering problem is also addressed by a binning approach seamlessly built into the convolution. This convolution operator then serves as the basic building block of a neural network that is robust to point clouds under 6-DoF transformations such as translation and rotation. Our experiment shows that our method performs with high accuracy in common scene understanding tasks such as object classification and segmentation. Compared to previous and concurrent works, most importantly, our method is able to generalize and achieve consistent results across different scenarios in which training and testing can contain arbitrary rotations. Our implementation is publicly available at our project page 1 .

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Rotation Invariant Convolutions for 3D Point Clouds Deep Learning

Zhang

Hua

Rosen

et al. 2019

2019 International Conference on 3D Vision (3DV)

Self Cite

136

144

View full text Add to dashboard Cite

show abstract

“…More recently, deep learning has increasingly gained attention in various applications. For instance, 3D neural networks have been extensively explored for 3D object detection and reconstruction (Zhi et al, 2018), 3D semantic segmentation (Engelmann et al, 2017;Graham et al, 2017;Tchapmi et al, 2017), 3D classification of point clouds (Özdemir et al, 2019;Uy et al, 2019) and 3D object pose estimation (Qi et al, 2019).…”

Section: * Corresponding Authormentioning

confidence: 99%

Holistic Parametric Reconstruction of Building Models From Point Clouds

Li¹,

Zhang

Shan³

2020

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. Building models are conventionally reconstructed by building roof points via planar segmentation and then using a topology graph to group the planes together. Roof edges and vertices are then mathematically represented by intersecting segmented planes. Technically, such solution is based on sequential local fitting, i.e., the entire data of one building are not simultaneously participating in determining the building model. As a consequence, the solution is lack of topological integrity and geometric rigor. Fundamentally different from this traditional approach, we propose a holistic parametric reconstruction method which means taking into consideration the entire point clouds of one building simultaneously. In our work, building models are reconstructed from predefined parametric (roof) primitives. We first use a well-designed deep neural network to segment and identify primitives in the given building point clouds. A holistic optimization strategy is then introduced to simultaneously determine the parameters of a segmented primitive. In the last step, the optimal parameters are used to generate a watertight building model in CityGML format. The airborne LiDAR dataset RoofN3D with predefined roof types is used for our test. It is shown that PointNet++ applied to the entire dataset can achieve an accuracy of 83% for primitive classification. For a subset of 910 buildings in RoofN3D, the holistic approach is then used to determine the parameters of primitives and reconstruct the buildings. The achieved overall quality of reconstruction is 0.08 meters for point-surface-distance or 0.7 times RMSE of the input LiDAR points. This study demonstrates the efficiency and capability of the proposed approach and its potential to handle large scale urban point clouds.

show abstract

“…However, several scholars have developed effective solutions based on deep learning to solve the problem of point cloud classification [3], [13], [14]. Nevertheless, as pointed out by Uy et al [30], many of the methods proposed for point cloud classification are successful with synthetic data but experience problems when applied to realistic data containing much noise and complex background information.…”

Section: Introductionmentioning

confidence: 99%

Point Cloud Classification Model Based on a Dual-Input Deep Network Framework

Zhai

Guo

et al. 2020

IEEE Access

View full text Add to dashboard Cite

The disorder, sparseness, irregularity, noise, and background of point clouds cause significant challenges in point cloud classification tasks. In such tasks, deep learning methods based on raw point cloud data have recently achieved good performance on simulated data. However, many methods experience problems when applied to realistic data containing much noise and complex background information. This paper proposes an end-to-end dual-input network (DINet) point cloud classification model based on deep learning. In the proposed model, a feature extractor obtains high-dimensional features, a feature comparator aggregates and disperses homogeneous and heterogeneous point clouds, respectively, in the feature space, and a feature analyzer completes the task. The two-channel data input facilitates a universal DINet framework that is flexible and extends to other models generalized across different datasets. DINet brings improvements in performance, presenting an overall accuracy value of 81.3% and average accuracy value of 79.6% in experiments conducted on the real-world ScanObjectNN dataset. The code for the proposed point cloud classification model is available at https://github.com/zhairf/DINet. INDEX TERMS 3D point cloud, neural networks, regularization, semantic classification.

show abstract

Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data

Cited by 491 publications

References 41 publications

Rotation Invariant Convolutions for 3D Point Clouds Deep Learning

Rotation Invariant Convolutions for 3D Point Clouds Deep Learning

Holistic Parametric Reconstruction of Building Models From Point Clouds

Point Cloud Classification Model Based on a Dual-Input Deep Network Framework

Contact Info

Product

Resources

About