Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentation

By Xinge Zhu, Hui Zhou, Tai Wang, Fangzhou Hong, Yuexin Ma, Wei Li, Hongsheng Li, Dahua Lin. Computer Vision and Pattern Recognition, 2020

723 citations recorded by the source corpus.

State-of-the-art methods for large-scale driving-scene LiDAR segmentation often project the point clouds to 2D space and then process them via 2D convolution. Although this corporation shows the competitiveness in the point cloud, it inevitably alters and abandons the 3D topology and geometric relations. A natural remedy is to utilize the 3D voxelization and 3D convolution network. However, we found that in the outdoor point cloud, the improvement obtained in this way is quite limited. An important reason is the property of the outdoor point cloud, namely sparsity and varying density. Motivated by this investigation, we propose a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern while maintaining these inherent properties. Moreover, a point-wise refinement module is introduced to alleviate the interference of lossy voxel-based label encoding. We evaluate the proposed model on two large-scale datasets, i.e., SemanticKITTI and nuScenes. Our method achieves the 1st place in the leaderboard of SemanticKITTI and outperforms existing methods on nuScenes with a noticeable margin, about 4%. Furthermore, the proposed 3D framework also generalizes well to LiDAR panoptic segmentation and LiDAR 3D detection.

Subfields: Aerospace Engineering, Computational Mechanics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/bf872d5e3fcf46f44a1b9dd76ee0db9a8cc98e06

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Jiangmiao Pang
2. Song Han
3. Xinge Zhu
4. Tai Wang
5. Dahua Lin
6. Zhijian Liu
7. Yunpeng Zhang
8. Zhengbiao Zhu

In Computational Mechanics

1. Zhengbiao Zhu
2. Dalong Du
3. Yunpeng Zhang
4. Song Han
5. Daniela Rus
6. Haotian Tang
7. Jiwen Lu
8. Chen Change Loy

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Chen Change Loy
3. Dacheng Tao
4. Xiaogang Wang
5. Yi Yang
6. Trevor Darrell
7. Dahua Lin
8. Jie Zhou

Citations in the corpus

• RPVNet: A Deep and Efficient Range-Point-Voxel Fusion Network for LiDAR Point Cloud Segmentation (2021, background): “1∗ - - - - - - - - - - - - - - - - - - Cylinder3D [44] 67.”
• OccFormer: Dual-path Transformer for Vision-based 3D Semantic Occupancy Prediction (2023, background): “Though LiDAR-based methods [23, 44, 62, 43, 39, 53], with explicit depth measurements, have been dominating the leading performance on public datasets [15, 3, 46, 2], vision-based approaches still offer advantages in terms of cost-effectiveness, stability, and generality.”
• Probabilistic and Geometric Depth: Detecting Objects in Perspective (2021, positive): “Recent advanced methods in this field typically resort to various sensors, such as LiDAR [1, 2, 3, 4, 5, 6], Radar [7], binocular vision [8, 9], or their combinations for accurate depth information.”
• BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation (2022, methodology): “There are also U-Net like models specialized for 3D semantic segmentation [17, 13, 52, 33, 75], an important task for offline HD map construction.”
• Unsupervised Point Cloud Pre-training via Occlusion Completion (2020, background): “There has been a flurry of exciting new point cloud models for object detection [27, 52, 64] and segmentation [22, 26, 57, 65].”
• FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection (2021, methodology): “Most state-of-the-art methods [38, 14, 29, 32, 40, 41] rely on the accurate 3D information provided by LiDAR point clouds, but it is a heavy burden to install expensive LiDARs on each vehicle.”
• Transformer-Based Visual Segmentation: A Survey (2023, positive): “Likepanopticsegmentation,most3Dpanoptic segmentationmethods[109],[110],[111],[112],[113]first predictsemanticsegmentationresults,separateinstancesbased onthesepredictionsandfusethetworesultstoobtainthefinal results.”
• SurroundOcc: Multi-Camera 3D Occupancy Prediction for Autonomous Driving (2023, background): “Also, it is flexible to extend to other 3D down-stream tasks such as 3D semantic segmentation [69, 73, 15].”