Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lin, Yuqiang, Chen, Kehua, Lockyer, Sam, Yadav, Arjun, Sui, Mingxuan, Zhang, Shucheng, Shi, Yan, Wang, Bingzhang, Zhang, Yuang, Zarbock, Markus, Stanek, Florain, Evans, Adrian, Li, Wenbin, Wang, Yinhai, Zhang, Nic
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2603.19098
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908901464604672
author Lin, Yuqiang
Chen, Kehua
Lockyer, Sam
Yadav, Arjun
Sui, Mingxuan
Zhang, Shucheng
Shi, Yan
Wang, Bingzhang
Zhang, Yuang
Zarbock, Markus
Stanek, Florain
Evans, Adrian
Li, Wenbin
Wang, Yinhai
Zhang, Nic
author_facet Lin, Yuqiang
Chen, Kehua
Lockyer, Sam
Yadav, Arjun
Sui, Mingxuan
Zhang, Shucheng
Shi, Yan
Wang, Bingzhang
Zhang, Yuang
Zarbock, Markus
Stanek, Florain
Evans, Adrian
Li, Wenbin
Wang, Yinhai
Zhang, Nic
contents Traffic Anomaly Understanding (TAU) is important for traffic safety in Intelligent Transportation Systems. Recent vision-language models (VLMs) have shown strong capabilities in video understanding. However, progress on TAU remains limited due to the lack of benchmarks and task-specific methodologies. To address this limitation, we introduce Roundabout-TAU, a dataset constructed from real-world roundabout videos collected in collaboration with the City of Carmel, Indiana. The dataset contains 342 clips and is annotated with more than 2,000 question-answer pairs covering multiple aspects of traffic anomaly understanding. Building on this benchmark, we propose TAU-R1, a two-layer vision-language framework for TAU. The first layer is a lightweight anomaly classifier that performs coarse anomaly categorisation, while the second layer is a larger anomaly reasoner that generates detailed event summaries. To improve task-specific reasoning, we introduce a two-stage training strategy consisting of decomposed-QA-enhanced supervised fine-tuning followed by TAU-GRPO, a GRPO-based post-training method with TAU-specific reward functions. Experimental results show that TAU-R1 achieves strong performance on both anomaly classification and reasoning tasks while maintaining deployment efficiency. The dataset and code are available at: https://github.com/siri-rouser/TAU-R1
format Preprint
id arxiv_https___arxiv_org_abs_2603_19098
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TAU-R1: Visual Language Model for Traffic Anomaly Understanding
Lin, Yuqiang
Chen, Kehua
Lockyer, Sam
Yadav, Arjun
Sui, Mingxuan
Zhang, Shucheng
Shi, Yan
Wang, Bingzhang
Zhang, Yuang
Zarbock, Markus
Stanek, Florain
Evans, Adrian
Li, Wenbin
Wang, Yinhai
Zhang, Nic
Computer Vision and Pattern Recognition
Traffic Anomaly Understanding (TAU) is important for traffic safety in Intelligent Transportation Systems. Recent vision-language models (VLMs) have shown strong capabilities in video understanding. However, progress on TAU remains limited due to the lack of benchmarks and task-specific methodologies. To address this limitation, we introduce Roundabout-TAU, a dataset constructed from real-world roundabout videos collected in collaboration with the City of Carmel, Indiana. The dataset contains 342 clips and is annotated with more than 2,000 question-answer pairs covering multiple aspects of traffic anomaly understanding. Building on this benchmark, we propose TAU-R1, a two-layer vision-language framework for TAU. The first layer is a lightweight anomaly classifier that performs coarse anomaly categorisation, while the second layer is a larger anomaly reasoner that generates detailed event summaries. To improve task-specific reasoning, we introduce a two-stage training strategy consisting of decomposed-QA-enhanced supervised fine-tuning followed by TAU-GRPO, a GRPO-based post-training method with TAU-specific reward functions. Experimental results show that TAU-R1 achieves strong performance on both anomaly classification and reasoning tasks while maintaining deployment efficiency. The dataset and code are available at: https://github.com/siri-rouser/TAU-R1
title TAU-R1: Visual Language Model for Traffic Anomaly Understanding
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.19098