SECOND: Sparsely Embedded Convolutional Detection
LiDAR-based or RGB-D-based object detection is used in numerous applications, ranging from autonomous driving to robot vision. Voxel-based 3D convolutional networks have been used for some time to enhance the retention of information when processing point cloud LiDAR data. However, problems remain,...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2018-10-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | http://www.mdpi.com/1424-8220/18/10/3337 |
| _version_ | 1818040603508736000 |
|---|---|
| author | Yan Yan Yuxing Mao Bo Li |
| author_facet | Yan Yan Yuxing Mao Bo Li |
| author_sort | Yan Yan |
| collection | DOAJ |
| description | LiDAR-based or RGB-D-based object detection is used in numerous applications, ranging from autonomous driving to robot vision. Voxel-based 3D convolutional networks have been used for some time to enhance the retention of information when processing point cloud LiDAR data. However, problems remain, including a slow inference speed and low orientation estimation performance. We therefore investigate an improved sparse convolution method for such networks, which significantly increases the speed of both training and inference. We also introduce a new form of angle loss regression to improve the orientation estimation performance and a new data augmentation approach that can enhance the convergence speed and performance. The proposed network produces state-of-the-art results on the KITTI 3D object detection benchmarks while maintaining a fast inference speed. |
| first_indexed | 2024-12-10T08:17:09Z |
| format | Article |
| id | doaj.art-53a182b0009f4a1696d217c4a0a01249 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-12-10T08:17:09Z |
| publishDate | 2018-10-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-53a182b0009f4a1696d217c4a0a012492022-12-22T01:56:27ZengMDPI AGSensors1424-82202018-10-011810333710.3390/s18103337s18103337SECOND: Sparsely Embedded Convolutional DetectionYan Yan0Yuxing Mao1Bo Li2State Key Laboratory of Power Transmission Equipment and System Security and New Technology, Chongqing University, Chongqing 400044, ChinaState Key Laboratory of Power Transmission Equipment and System Security and New Technology, Chongqing University, Chongqing 400044, ChinaTrunkTech Co., Ltd., No. 3, Danling street, ZhongGuan Town, HaiDian District, Beijing 100089, ChinaLiDAR-based or RGB-D-based object detection is used in numerous applications, ranging from autonomous driving to robot vision. Voxel-based 3D convolutional networks have been used for some time to enhance the retention of information when processing point cloud LiDAR data. However, problems remain, including a slow inference speed and low orientation estimation performance. We therefore investigate an improved sparse convolution method for such networks, which significantly increases the speed of both training and inference. We also introduce a new form of angle loss regression to improve the orientation estimation performance and a new data augmentation approach that can enhance the convergence speed and performance. The proposed network produces state-of-the-art results on the KITTI 3D object detection benchmarks while maintaining a fast inference speed.http://www.mdpi.com/1424-8220/18/10/33373D object detectionconvolutional neural networksLIDARautonomous driving |
| spellingShingle | Yan Yan Yuxing Mao Bo Li SECOND: Sparsely Embedded Convolutional Detection Sensors 3D object detection convolutional neural networks LIDAR autonomous driving |
| title | SECOND: Sparsely Embedded Convolutional Detection |
| title_full | SECOND: Sparsely Embedded Convolutional Detection |
| title_fullStr | SECOND: Sparsely Embedded Convolutional Detection |
| title_full_unstemmed | SECOND: Sparsely Embedded Convolutional Detection |
| title_short | SECOND: Sparsely Embedded Convolutional Detection |
| title_sort | second sparsely embedded convolutional detection |
| topic | 3D object detection convolutional neural networks LIDAR autonomous driving |
| url | http://www.mdpi.com/1424-8220/18/10/3337 |
| work_keys_str_mv | AT yanyan secondsparselyembeddedconvolutionaldetection AT yuxingmao secondsparselyembeddedconvolutionaldetection AT boli secondsparselyembeddedconvolutionaldetection |