Few-shot learning for novel object detection in autonomous driving
Artificial intelligence and advanced sensing technologies have significantly advanced the intelligent transportation system and autonomous vehicles. Perception, a critical component, extracts real-time traffic information essential for various system functionalities, such as agent behavior predictio...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Communications in Transportation Research |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2772424725000344 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849317677876641792 |
|---|---|
| author | Yifan Zhuang Pei Liu Hao Yang Kai Zhang Yinhai Wang Ziyuan Pu |
| author_facet | Yifan Zhuang Pei Liu Hao Yang Kai Zhang Yinhai Wang Ziyuan Pu |
| author_sort | Yifan Zhuang |
| collection | DOAJ |
| description | Artificial intelligence and advanced sensing technologies have significantly advanced the intelligent transportation system and autonomous vehicles. Perception, a critical component, extracts real-time traffic information essential for various system functionalities, such as agent behavior prediction. However, the quality of information derived from perception greatly influences overall system performance. This study focuses on enhancing perception robustness in autonomous vehicles, particularly in detecting rare objects, which pose a challenge due to limited training samples. While deep learning-based vision methods have shown promising accuracy, they struggle with rare object detection. To address this, we propose a few-shot learning training strategy tailored for improved detection accuracy of rare or novel objects. Additionally, we design a one-stage object detector for efficient object detection in autonomous driving scenarios. Experiments on a self-driving dataset augmented with rare objects alongside the popular few-shot object detection (FSOD) benchmark, the pattern analysis, statical modeling, and computational learning PASCAL Visual Object Classes (PASCAL-VOC), demonstrate state-of-the-art accuracy in rare categories and superior inference speed compared to alternative algorithms. Furthermore, we investigate the impact of intra-class variance on detection accuracy, providing insights for data annotation in the preparation stage. |
| format | Article |
| id | doaj-art-b6fee2bb24904d30a05ff40c7d988aed |
| institution | Kabale University |
| issn | 2772-4247 |
| language | English |
| publishDate | 2025-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Communications in Transportation Research |
| spelling | doaj-art-b6fee2bb24904d30a05ff40c7d988aed2025-08-20T03:51:08ZengElsevierCommunications in Transportation Research2772-42472025-12-01510019410.1016/j.commtr.2025.100194Few-shot learning for novel object detection in autonomous drivingYifan Zhuang0Pei Liu1Hao Yang2Kai Zhang3Yinhai Wang4Ziyuan Pu5Department of Civil and Environmental Engineering, University of Washington, Seattle, 98195, USAIntelligent Transportation Thrust, Systems Hub, The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 511453, ChinaDepartment of Civil and Environmental Engineering, University of Washington, Seattle, 98195, USATsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, 518055, ChinaDepartment of Civil and Environmental Engineering, University of Washington, Seattle, 98195, USASchool of Transportation, Southeast University, Nanjing, 211102, China; Corresponding author.Artificial intelligence and advanced sensing technologies have significantly advanced the intelligent transportation system and autonomous vehicles. Perception, a critical component, extracts real-time traffic information essential for various system functionalities, such as agent behavior prediction. However, the quality of information derived from perception greatly influences overall system performance. This study focuses on enhancing perception robustness in autonomous vehicles, particularly in detecting rare objects, which pose a challenge due to limited training samples. While deep learning-based vision methods have shown promising accuracy, they struggle with rare object detection. To address this, we propose a few-shot learning training strategy tailored for improved detection accuracy of rare or novel objects. Additionally, we design a one-stage object detector for efficient object detection in autonomous driving scenarios. Experiments on a self-driving dataset augmented with rare objects alongside the popular few-shot object detection (FSOD) benchmark, the pattern analysis, statical modeling, and computational learning PASCAL Visual Object Classes (PASCAL-VOC), demonstrate state-of-the-art accuracy in rare categories and superior inference speed compared to alternative algorithms. Furthermore, we investigate the impact of intra-class variance on detection accuracy, providing insights for data annotation in the preparation stage.http://www.sciencedirect.com/science/article/pii/S2772424725000344Few-shot learningComputer visionAutonomous drivingObject detection |
| spellingShingle | Yifan Zhuang Pei Liu Hao Yang Kai Zhang Yinhai Wang Ziyuan Pu Few-shot learning for novel object detection in autonomous driving Communications in Transportation Research Few-shot learning Computer vision Autonomous driving Object detection |
| title | Few-shot learning for novel object detection in autonomous driving |
| title_full | Few-shot learning for novel object detection in autonomous driving |
| title_fullStr | Few-shot learning for novel object detection in autonomous driving |
| title_full_unstemmed | Few-shot learning for novel object detection in autonomous driving |
| title_short | Few-shot learning for novel object detection in autonomous driving |
| title_sort | few shot learning for novel object detection in autonomous driving |
| topic | Few-shot learning Computer vision Autonomous driving Object detection |
| url | http://www.sciencedirect.com/science/article/pii/S2772424725000344 |
| work_keys_str_mv | AT yifanzhuang fewshotlearningfornovelobjectdetectioninautonomousdriving AT peiliu fewshotlearningfornovelobjectdetectioninautonomousdriving AT haoyang fewshotlearningfornovelobjectdetectioninautonomousdriving AT kaizhang fewshotlearningfornovelobjectdetectioninautonomousdriving AT yinhaiwang fewshotlearningfornovelobjectdetectioninautonomousdriving AT ziyuanpu fewshotlearningfornovelobjectdetectioninautonomousdriving |