Filters








6 Hits in 1.4 sec

HVNet: Hybrid Voxel Network for LiDAR Based 3D Object Detection [article]

Maosheng Ye, Shuangjie Xu, Tongyi Cao
2020 arXiv   pre-print
We present Hybrid Voxel Network (HVNet), a novel one-stage unified network for point cloud based 3D object detection for autonomous driving.  ...  Recent studies show that 2D voxelization with per voxel PointNet style feature extractor leads to accurate and efficient detector for large 3D scenes.  ...  Related Work 3D Object Detection There are roughly two different lines for existing methods of 3D object detection with point cloud: Multi-sensor based 3D object detection.  ... 
arXiv:2003.00186v2 fatcat:lzcfuqga3vhx3hn2s6r2jmz2ua

PillarNet: Real-Time and High-Performance Pillar-based 3D Object Detection [article]

Guangsheng Shi, Ruifeng Li, Chao Ma
2022 arXiv   pre-print
Real-time and high-performance 3D object detection is of critical importance for autonomous driving.  ...  Recent top-performing 3D object detectors mainly rely on point-based or 3D voxel-based convolutions, which are both computationally inefficient for onboard deployment.  ...  The hybrid 3D representation keeps the efficiency of pillar-based detection while implicitly leveraging the voxel-based feature learning regime.  ... 
arXiv:2205.07403v2 fatcat:7hpppd6r7jeajp3vfyqj4octuu

Behind the Curtain: Learning Occluded Shapes for 3D Object Detection [article]

Qiangeng Xu, Yiqi Zhong, Ulrich Neumann
2021 arXiv   pre-print
To tackle the challenge, we present a novel LiDAR-based 3D object detection model, dubbed Behind the Curtain Detector (BtcDet), which learns the object shape priors and estimates the complete object shapes  ...  Particularly, for the 3D detection of both cars and cyclists on the KITTI benchmark, BtcDet surpasses all of the published state-of-the-art methods by remarkable margins.  ...  Hvnet: Hybrid voxel net- Shi, S.; Wang, X.; and Li, H. 2019a. Pointrcnn: 3d object work for lidar based 3d object detection.  ... 
arXiv:2112.02205v1 fatcat:37bxvqldfzdbdeklzjvxzgkaou

Voxel Transformer for 3D Object Detection [article]

Jiageng Mao and Yujing Xue and Minzhe Niu and Haoyue Bai and Jiashi Feng and Xiaodan Liang and Hang Xu and Chunjing Xu
2021 arXiv   pre-print
We present Voxel Transformer (VoTr), a novel and effective voxel-based Transformer backbone for 3D object detection from point clouds.  ...  Conventional 3D convolutional backbones in voxel-based 3D detectors cannot efficiently capture large context information, which is crucial for object recognition and localization, owing to the limited  ...  HVNet [37] designs a convolutional network that leverages the hybrid voxel representation. PV-RCNN [26] uses keypoints to extract voxel features for boxes refinement.  ... 
arXiv:2109.02497v2 fatcat:ouryanxvifacnniqloydc3r5gq

From Multi-View to Hollow-3D: Hallucinated Hollow-3D R-CNN for 3D Object Detection [article]

Jiajun Deng, Wengang Zhou, Yanyong Zhang, Houqiang Li
2021 arXiv   pre-print
Finally, the 3D objects are detected via a box refinement module with a novel Hierarchical Voxel RoI Pooling operation.  ...  However, point clouds are always sparsely distributed in the 3D space, and with unstructured storage, which makes it difficult to represent them for effective 3D object detection.  ...  The resulting hollow-3D representation is applicable in scene understanding with LiDAR point clouds, and in this paper we focus on region-based 3D object detection.  ... 
arXiv:2107.14391v1 fatcat:ucy3rhray5dznntzb6y47ierxu

DRINet: A Dual-Representation Iterative Learning Network for Point Cloud Segmentation [article]

Maosheng Ye, Shuangjie Xu, Tongyi Cao, Qifeng Chen
2021 arXiv   pre-print
DRINet mainly consists of two modules called Sparse Point-Voxel Feature Extraction and Sparse Voxel-Point Feature Extraction.  ...  Our network achieves state-of-the-art results for point cloud classification and segmentation tasks on several datasets while maintaining high runtime efficiency.  ...  Most works in lidar detection and segmentation [17, 47] adopt this way to form 2D/3D birdeye view image features.  ... 
arXiv:2108.04023v1 fatcat:q6dkzjhikzcbxjhqno2ee7jwwq