An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving
The environment perception algorithm in autonomous driving is trained in the source domain, leading to domain drift and reduced detection accuracy in the target domain due to shifts in background feature distribution. To address this issue, a domain adaptive object detection algorithm based on featu...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/13/11/6448 |
| _version_ | 1827739790711193600 |
|---|---|
| author | Yuan Zhu Ruidong Xu Chongben Tao Hao An Zhipeng Sun Ke Lu |
| author_facet | Yuan Zhu Ruidong Xu Chongben Tao Hao An Zhipeng Sun Ke Lu |
| author_sort | Yuan Zhu |
| collection | DOAJ |
| description | The environment perception algorithm in autonomous driving is trained in the source domain, leading to domain drift and reduced detection accuracy in the target domain due to shifts in background feature distribution. To address this issue, a domain adaptive object detection algorithm based on feature uncertainty is proposed, which can improve the detection performance of object detection algorithms in unlabeled data. Firstly, a local alignment module based on channel information is proposed, which can obtain the model’s uncertainty about different domain data based on the feature channels obtained through the feature extraction network, achieving adaptive dynamic local alignment. Secondly, an instance-level alignment module guided by local feature uncertainty is proposed, which can obtain the corresponding instance-level uncertainty through ROI mapping. To improve the domain invariance of bounding box regression, a multi-class, multi-regression instance-level uncertainty alignment module is proposed, which can achieve spatial decoupling of classification and regression tasks, further improving the model’s domain adaptive ability. Finally, the effectiveness of the proposed algorithm is validated on Cityscapes, KITTI, and real vehicle data. |
| first_indexed | 2024-03-11T03:13:03Z |
| format | Article |
| id | doaj.art-fb0375ae6b7c484ab3b1bb41bb6a8cdf |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-11T03:13:03Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-fb0375ae6b7c484ab3b1bb41bb6a8cdf2023-11-18T07:32:16ZengMDPI AGApplied Sciences2076-34172023-05-011311644810.3390/app13116448An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous DrivingYuan Zhu0Ruidong Xu1Chongben Tao2Hao An3Zhipeng Sun4Ke Lu5School of Automotive Studies, Tongji University, Shanghai 201800, ChinaSchool of Automotive Studies, Tongji University, Shanghai 201800, ChinaThe School of Electronic and Information Engineering, Suzhou University of Science and Technology, Suzhou 215009, ChinaSchool of Automotive Studies, Tongji University, Shanghai 201800, ChinaNanchang Automotive Institute of Intelligence & New Energy, Tongji University, Nanchang 330013, ChinaSchool of Automotive Studies, Tongji University, Shanghai 201800, ChinaThe environment perception algorithm in autonomous driving is trained in the source domain, leading to domain drift and reduced detection accuracy in the target domain due to shifts in background feature distribution. To address this issue, a domain adaptive object detection algorithm based on feature uncertainty is proposed, which can improve the detection performance of object detection algorithms in unlabeled data. Firstly, a local alignment module based on channel information is proposed, which can obtain the model’s uncertainty about different domain data based on the feature channels obtained through the feature extraction network, achieving adaptive dynamic local alignment. Secondly, an instance-level alignment module guided by local feature uncertainty is proposed, which can obtain the corresponding instance-level uncertainty through ROI mapping. To improve the domain invariance of bounding box regression, a multi-class, multi-regression instance-level uncertainty alignment module is proposed, which can achieve spatial decoupling of classification and regression tasks, further improving the model’s domain adaptive ability. Finally, the effectiveness of the proposed algorithm is validated on Cityscapes, KITTI, and real vehicle data.https://www.mdpi.com/2076-3417/13/11/6448object detectiondomain adaptationuncertainty |
| spellingShingle | Yuan Zhu Ruidong Xu Chongben Tao Hao An Zhipeng Sun Ke Lu An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving Applied Sciences object detection domain adaptation uncertainty |
| title | An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving |
| title_full | An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving |
| title_fullStr | An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving |
| title_full_unstemmed | An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving |
| title_short | An Object Detection Method Based on Feature Uncertainty Domain Adaptation for Autonomous Driving |
| title_sort | object detection method based on feature uncertainty domain adaptation for autonomous driving |
| topic | object detection domain adaptation uncertainty |
| url | https://www.mdpi.com/2076-3417/13/11/6448 |
| work_keys_str_mv | AT yuanzhu anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT ruidongxu anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT chongbentao anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT haoan anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT zhipengsun anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT kelu anobjectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT yuanzhu objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT ruidongxu objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT chongbentao objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT haoan objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT zhipengsun objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving AT kelu objectdetectionmethodbasedonfeatureuncertaintydomainadaptationforautonomousdriving |