LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection
Lightweight modules play a key role in 3D object detection tasks for autonomous driving, which are necessary for the application of 3D object detectors. At present, research still focuses on constructing complex models and calculations to improve the detection precision at the expense of the running...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Tsinghua University Press
2022-12-01
|
| Series: | CAAI Artificial Intelligence Research |
| Subjects: | |
| Online Access: | https://www.sciopen.com/article/10.26599/AIR.2022.9150009 |
| _version_ | 1797293814284550144 |
|---|---|
| author | Shuo Yang Huimin Lu Tohru Kamiya Yoshihisa Nakatoh Seiichi Serikawa |
| author_facet | Shuo Yang Huimin Lu Tohru Kamiya Yoshihisa Nakatoh Seiichi Serikawa |
| author_sort | Shuo Yang |
| collection | DOAJ |
| description | Lightweight modules play a key role in 3D object detection tasks for autonomous driving, which are necessary for the application of 3D object detectors. At present, research still focuses on constructing complex models and calculations to improve the detection precision at the expense of the running rate. However, building a lightweight model to learn the global features from point cloud data for 3D object detection is a significant problem. In this paper, we focus on combining convolutional neural networks with self-attention-based vision transformers to realize lightweight and high-speed computing for 3D object detection. We propose light-weight detection 3D (LWD-3D), which is a point cloud conversion and lightweight vision transformer for autonomous driving. LWD-3D utilizes a one-shot regression framework in 2D space and generates a 3D object bounding box from point cloud data, which provides a new feature representation method based on a vision transformer for 3D detection applications. The results of experiment on the KITTI 3D dataset show that LWD-3D achieves real-time detection (time per image < 20 ms). LWD-3D obtains a mean average precision (mAP) 75% higher than that of another 3D real-time detector with half the number of parameters. Our research extends the application of visual transformers to 3D object detection tasks. |
| first_indexed | 2024-03-07T21:20:20Z |
| format | Article |
| id | doaj.art-c3363fba83604b6b9c05e1b308f5df50 |
| institution | Directory Open Access Journal |
| issn | 2097-194X |
| language | English |
| last_indexed | 2024-03-07T21:20:20Z |
| publishDate | 2022-12-01 |
| publisher | Tsinghua University Press |
| record_format | Article |
| series | CAAI Artificial Intelligence Research |
| spelling | doaj.art-c3363fba83604b6b9c05e1b308f5df502024-02-27T14:40:54ZengTsinghua University PressCAAI Artificial Intelligence Research2097-194X2022-12-011213714310.26599/AIR.2022.9150009LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object DetectionShuo Yang0Huimin Lu1Tohru Kamiya2Yoshihisa Nakatoh3Seiichi Serikawa4School of Engineering, Kyushu Institute of Technology, Fukuoka 804-8550, JapanSchool of Engineering, Kyushu Institute of Technology, Fukuoka 804-8550, JapanSchool of Engineering, Kyushu Institute of Technology, Fukuoka 804-8550, JapanSchool of Engineering, Kyushu Institute of Technology, Fukuoka 804-8550, JapanSchool of Engineering, Kyushu Institute of Technology, Fukuoka 804-8550, JapanLightweight modules play a key role in 3D object detection tasks for autonomous driving, which are necessary for the application of 3D object detectors. At present, research still focuses on constructing complex models and calculations to improve the detection precision at the expense of the running rate. However, building a lightweight model to learn the global features from point cloud data for 3D object detection is a significant problem. In this paper, we focus on combining convolutional neural networks with self-attention-based vision transformers to realize lightweight and high-speed computing for 3D object detection. We propose light-weight detection 3D (LWD-3D), which is a point cloud conversion and lightweight vision transformer for autonomous driving. LWD-3D utilizes a one-shot regression framework in 2D space and generates a 3D object bounding box from point cloud data, which provides a new feature representation method based on a vision transformer for 3D detection applications. The results of experiment on the KITTI 3D dataset show that LWD-3D achieves real-time detection (time per image < 20 ms). LWD-3D obtains a mean average precision (mAP) 75% higher than that of another 3D real-time detector with half the number of parameters. Our research extends the application of visual transformers to 3D object detection tasks.https://www.sciopen.com/article/10.26599/AIR.2022.91500093d object detectionpoint cloudsvision transformerone-shot regressionreal-time |
| spellingShingle | Shuo Yang Huimin Lu Tohru Kamiya Yoshihisa Nakatoh Seiichi Serikawa LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection CAAI Artificial Intelligence Research 3d object detection point clouds vision transformer one-shot regression real-time |
| title | LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection |
| title_full | LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection |
| title_fullStr | LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection |
| title_full_unstemmed | LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection |
| title_short | LWD-3D: Lightweight Detector Based on Self-Attention for 3D Object Detection |
| title_sort | lwd 3d lightweight detector based on self attention for 3d object detection |
| topic | 3d object detection point clouds vision transformer one-shot regression real-time |
| url | https://www.sciopen.com/article/10.26599/AIR.2022.9150009 |
| work_keys_str_mv | AT shuoyang lwd3dlightweightdetectorbasedonselfattentionfor3dobjectdetection AT huiminlu lwd3dlightweightdetectorbasedonselfattentionfor3dobjectdetection AT tohrukamiya lwd3dlightweightdetectorbasedonselfattentionfor3dobjectdetection AT yoshihisanakatoh lwd3dlightweightdetectorbasedonselfattentionfor3dobjectdetection AT seiichiserikawa lwd3dlightweightdetectorbasedonselfattentionfor3dobjectdetection |