Deep Learning for 3D Point Clouds: A Survey
By Yulan Guo, Hanyun Wang, Qingyong Hu, Hao Liu, Li Liu, Mohammed Bennamoun. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019
2,282 citations recorded by the source corpus.
Point cloud learning has lately attracted increasing attention due to its wide applications in many areas, such as computer vision, autonomous driving, and robotics. As a dominating technique in AI, deep learning has been successfully used to solve various 2D vision problems. However, deep learning on point clouds is still in its infancy due to the unique challenges faced by the processing of point clouds with deep neural networks. Recently, deep learning on point clouds has become even thriving, with numerous methods being proposed to address different problems in this area. To stimulate future research, this paper presents a comprehensive review of recent progress in deep learning methods for point clouds. It covers three major tasks, including 3D shape classification, 3D object detection and tracking, and 3D point cloud segmentation. It also presents comparative results on several publicly available datasets, together with insightful observations and inspiring future research directions.
Subfields: Computational Mechanics, Environmental Engineering, Geology
View Paper: https://www.semanticscholar.org/paper/0311ace1d499cadd1cc0c515a625d1d045f60d25
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computational Mechanics
- Shaoshuai Shi
- Xiaogang Wang
- Hongsheng Li
- Qingyong Hu
- Bruno E. Correia
- Freyr Sverrisson
- Yulan Guo
- Jiageng Mao
In Environmental Engineering
- Klaus C. J. Dietmayer
- Shaoshuai Shi
- Hongsheng Li
- Xiaogang Wang
- Qingyong Hu
- Jiageng Mao
- Yulan Guo
- Bo Li
In Geology
Citations in the corpus
- Revisiting Point Cloud Shape Classification with a Simple and Effective Baseline (2021, background): “For a detailed survey of various projection approaches on different point-cloud processing tasks, we encourage readers to check the recent survey paper by (Guo et al., 2020).”
- Semantic Segmentation on Swiss3DCities: A Benchmark Study on Aerial Photogrammetric 3D Pointcloud Dataset (2020, background): “…more information to discriminate between different classes; currently, state-of-the-art models for 3D semantic segmentation rely on deep learning [11], [39], [4] and represent input data as voxels [40], points [24] or meshes [17]; other approaches render multiple views of the 3D scene and then…”
- RELLIS-3D Dataset: Data, Benchmarks and Analysis (2020, methodology): “There are two types of deep learning methods for LiDAR Point Cloud: point-based methods and the projective method [34].”
- Invisible for both Camera and LiDAR: Security of Multi-Sensor Fusion based Perception in Autonomous Driving Under Physical-World Attacks (2021, background): “These pre-processing steps can largely reduce the sizes and dimensions of the MSF algorithm inputs, which can thus greatly improve the run-time algorithm performance [47].”
- Geometric Back-Projection Network for Point Cloud Classification (2019, methodology): “r 3D data usually incorporate geometry estimation and model reconstruction. Assisted by deep learning, recent works on 3D focus on data-driven approaches via Convolutional Neural Networks. Guo et al. [19] categorized CNN approaches to point clouds classification as: multi-view methods (e.g., MVCNN [2”
- MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion (2024, methodology): “During training, we sample the datasets asymmetrically to place extra weight on PointOdyssey (more dynamic, articulated objects) and less weight on TartanAir (good scene diversity but static) and Waymo (a highly specialized domain).”
- A Survey on Vision--Language--Action Models for Embodied AI (2024, neutral): “Point clouds [256] are also popular visual input types due to the widespread adoption of LiDARs and 3D scanners; depth maps can be easily converted to point clouds.”
- 3D Object Detection for Autonomous Driving: A Survey (2021, background): “As of this main text in June 2021, we notice that rather a few investigations [13, 35, 14, 36] have looked into this application issue.”
- Point Transformer (2020, background): “However, due to the irregularities of the shape and point density, point convolutions are usually hard to design and the kernel needs to be adapted for different input data [32].”
- Group-Free 3D Object Detection via Transformers (2021, background): “[13] provides a good taxonomy and review of all these architectures, and discussing all of them is beyond the scope of this paper.”
- SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (2020, background): “Accurate matching of partial 3D surfaces is critical for point cloud registration [17, 38, 6, 21, 29], segmentation [58, 28, 27], and recognition [24, 12, 47].”
- VoxFormer: Sparse Voxel Transformer for Camera-Based 3D Semantic Scene Completion (2023, background): “Human-level scene understanding for intelligent agents is typically advanced by semantic segmentation on images [32] or point clouds [33].”
- 3D Object Detection for Autonomous Driving: A Comprehensive Survey (2022, result): “In contrast to the previous surveys that only focus on detection from point cloud (Guo et al., 2020; Fernandes et al., 2021; Zamanakos et al., 2021), from monocular images (Wu et al.”
- Vision-based robotic grasping from object localization, object pose estimation to grasp estimation for parallel grippers: a review (2020, background): “sive review about 3D object detection could refer to the survey [Guo et al., 2020].”
- RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds (2019, background): “Recent learning based approaches [16, 43, 37] mainly include projection-based, voxel-based and pointbased schemes which are outlined here.”