TSBOW: Traffic Surveillance Benchmark
for Occluded Vehicles
Under Various Weather Conditions

Ngoc Doan-Minh Huynh   Duong Nguyen-Ngoc Tran  Long Hoang Pham

Tai Huu-Phuong Tran   Hyung-Joon Jeon   Huy-Hung Nguyen   Duong Khac Vu   Hyung-Min Jeon

Son Hong Phan   Quoc Pham-Nam Ho   Chi Dai Tran   Trinh Le Ba Khanh   Jae Wook Jeon

Automation Lab, Sungkyunkwan University

Corresponding Author: jwjeon@skku.edu
Contact for Dataset: {ngochdm, duongtran, phlong}@skku.edu

     
   

(UPDATING....)

(All links would be updated on the conference day.)

Please download our Github repo to get better markdown view (i.e. Visual Code).

🎉NEWS

• [2025.11.16] 🔥 Our code and website are released!
• [2025.11.08] 🎉 TSBOW has been accepted to AAAI 2026!

Abstract

Global warming has intensified the frequency and severity of extreme weather events, which degrade CCTV signal and video quality while disrupting traffic flow, thereby increasing traffic accident rates. Existing datasets, often limited to light haze, rain, and snow, fail to capture extreme weather conditions. To address this gap, this study introduces the Traffic Surveillance Benchmark for Occluded Vehicles under Various Weather Conditions (TSBOW), a comprehensive dataset designed to enhance occluded vehicle detection across diverse annual weather scenarios. Comprising over 32 hours of real-world traffic data from densely populated urban areas, TSBOW includes more than 48,000 manually annotated and 3.2 million semi-labeled frames; bounding boxes spanning eight traffic participant classes from large vehicles to micromobility devices and pedestrians. We establish an object detection benchmark for TSBOW, highlighting challenges posed by occlusions and adverse weather. With its varied road types, scales, and viewpoints, TSBOW serves as a critical resource for advancing Intelligent Transportation Systems. Our findings underscore the potential of CCTV-based traffic monitoring, paving the way for new research and applications. The TSBOW dataset is publicly available at the following link.
Code -- https://github.com/SKKUAutoLab/TSBOW

Overview

Dataset Statistics

Recording Locations

Video Distribution

Class Distribution

Other Distributions

Occlusion Ditribution

Traffic Distribution

Datasets

Comparison with other datasets

Comparison with other datasets about weather conditions and scales

Dataset	Introduction	Pub	Paper
UAVDT [website]	- Hardware: UAVs. - Tasks: object detection, single object tracking, multiple-object tracking. - Position: China. - Weather: sunny/cloudy, fog, rain. - Time: day, night.	IJCV 2020	The Unmanned Aerial Vehicle Benchmark: Object Detection, Tracking and Baseline
UA-DETRAC [website]	- Hardware: Cannon EOS 550D camera. - Tasks: object detection, multi-object tracking. - Position: China. - Weather: sunny/cloudy, rain. - Time: day, night.	CVIU 2020	UA-DETRAC: A new benchmark and protocol for multi-object detection and tracking
AAU RainSnow [website]	- Hardware: RGB color and thermal camera. - Tasks: instance segmentation, single object tracking, multiple-object tracking. - Position: Denmark. - Weather: fog, rain, snow. - Time: day, night.	ITS 2019	Rain Removal in Traffic Surveillance: Does it Matter?
TSBOW [website]	- Hardware: CCTV system + color camera. - Tasks: object detection. - Position: South Korea. - Weather: sunny/cloudy, haze, rain, snow. - Time: day. (night-time and other tasks will be updated later)	AAAI 2026	TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions

Baselines

Year	Pub	Paper	Link	Note
2024	ICDICI	A review on yolov8 and its advancements	paper	YOLOv8
2024	arXiV	YOLOv11: An Overview of the Key Architectural Enhancements	paper	YOLOv11
2024	CVPR	DETRs Beat YOLOs on Real-time Object Detection	paper	RT-DETR
2025	arXiV	A Breakdown of the Key Architectural Features	paper	YOLOv12

The source codes for baseline models are provided in Baselines folder. Read Instruction for more information.

Experiments

Model performances after training 100 epochs and validating with imgsz=1280 on manually labeled test set.

Model performances under different weather conditions

Comparison with other datasets

Model performances when training on different datasets

Comparison of traffic surveillance datasets

Models performance for car across different metrics on the comparison set

Ablation Studies

YOLOv12x performance across different classes.

Influence of dataset characteristics on object detection performance.

Dataset Download

(Upcoming) We will provide Terms and Conditions before downloading our TSBOW dataset.

Submission Guidelines

We will provide the guidelines soon.

(Upcoming) Scripts to download TSBOW from HuggingFace will be provided. Please refer to the download_TSBOW.py script for more details.

References

Thanks to the developers and contributors of the following open-source repositories, whose invaluable work has greatly inspire our project:

Datasets:

• UAVDT: A traffic dataset contains drone footages under sunny and rainy conditions.
• UA-DETRAC: A traffic surveillance dataset captures sunny and rainy weather.
• AAU RainSnow: A traffic surveillance dataset provides segmentation annotations for rain and snow weather.

Github Repo:

• X-AnyLabeling: An open-source tool for precise bounding box creation.
• Ultralytics YOLO: Detection models for training and real-time inferencing.
• YOLOv12: A model for object detection.

Our repository is licensed under the Apache 2.0 License. However, if you use other components in your work, please follow their license.

Citation

If our research is helpful to you, please cite our paper using the following BibTeX format

@article{Huynh_TSBOW_AAAI_2026,
    title   = {TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions},
    author  = {Ngoc Doan-Minh Huynh, Duong Nguyen-Ngoc Tran, Long Hoang Pham, Tai Huu-Phuong Tran, Hyung-Joon Jeon, Huy-Hung Nguyen, Duong Khac Vu, Hyung-Min Jeon, Son Hong Phan, Quoc Pham-Nam Ho, Chi Dai Tran, Trinh Le Ba Khanh, Jae Wook Jeon},
    journal = {AAAI 2026},
    year    = {2025}
}

🔝 Back to Top