README.md

---
task_categories:
- object-detection
- other
language:
- en
pretty_name: TSBOW
tags:
- traffic-surveillance
- benchmark-dataset
- object-detection
- image
- video
license: cc-by-nc-nd-4.0
gated: true
extra_gated_heading: "Acknowledge license (cc-by-nc-nd-4.0) to accept the repository"
# extra_gated_description: "This repository is publicly accessible, but you have to accept the conditions to access its files and content."
extra_gated_prompt: "The access request is only accepted after we receive your signed form via email. Our team may take 2-3 days to process your request."
extra_gated_button_content: "Agree and send request to access TSBOW"
---

<!-- 
__        _______ _     ____ ___  __  __ _____ 
\ \      / / ____| |   / ___/ _ \|  \/  | ____|
 \ \ /\ / /|  _| | |  | |  | | | | |\/| |  _|  
  \ V  V / | |___| |__| |__| |_| | |  | | |___ 
   \_/\_/  |_____|_____\____\___/|_|  |_|_____|
                                   
 _____ ___  
|_   _/ _ \ 
  | || | | |
  | || |_| |
  |_| \___/ 
            

 _____ ____  ____   _____        __
|_   _/ ___|| __ ) / _ \ \      / /
  | | \___ \|  _ \| | | \ \ /\ / / 
  | |  ___) | |_) | |_| |\ V  V /  
  |_| |____/|____/ \___/  \_/\_/   
                                   
-->


<h1 align='center'>
    TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles<br>
    Under Various Weather Conditions
</h1>


<!-- MARK: authors -->
<div align='center'>
    <a href="https://scholar.google.com/citations?user=pCTUkWwAAAAJ">
        Ngoc Doan-Minh Huynh</a> &emsp;
    <a href="https://scholar.google.com/citations?user=crRQGUAAAAAJ">
        Duong Nguyen-Ngoc Tran</a> &emsp;
    <a href="https://scholar.google.com/citations?user=xPyle9AAAAAJ">
        Long Hoang Pham</a>   
</div>

<div align='center'>
    Tai Huu-Phuong Tran &emsp;
    Hyung-Joon Jeon     &emsp;
    Huy-Hung Nguyen     &emsp;
    Duong Khac Vu       &emsp;
    Hyung-Min Jeon      
</div>

<div align='center'>
    Son Hong Phan       &emsp; 
    Quoc Pham-Nam Ho    &emsp;
    Chi Dai Tran        &emsp;
    Trinh Le Ba Khanh   &emsp;
    <a href="https://scholar.google.com/citations?user=9z0SfKoAAAAJ">
        Jae Wook Jeon</a>
</div>


<!-- affliation -->
<div align='center'>
    <a href="https://micro.skku.ac.kr/micro/index.do">Automation Lab</a>, Sungkyunkwan University, South Korea
</div>


<!-- contact -->
<div align='center'>
    <b>Corresponding Author</b>: jwjeon@skku.edu
    <br>
    <b>Contact for Dataset</b>: jwjeon@skku.edu, ngochdm@skku.edu, automation.skku@gmail.com
</div>


<!-- MARK: URLs -->
<!-- get img shields at: -->
<!-- https://shields.io/badges -->
<!-- check icon at: -->
<!-- https://github.com/simple-icons/simple-icons/blob/master/slugs.md  -->
<br> 

<div align="center">
  <a href="https://skkuautolab.github.io/TSBOW/"><img src="https://img.shields.io/static/v1?label=TSBOW&message=Website&color=9a33fc&logo=githubpages" style="height: 25px;"></a>  
  <a href="https://doi.org/10.1609/aaai.v40i7.37439"><img src="https://img.shields.io/static/v1?label=DOI&message=10.1609/aaai.v40i7.37439&color=green" style="height: 25px;"></a>  
  <a href="https://arxiv.org/abs/2602.05414"><img src="https://img.shields.io/static/v1?label=Supplementary&message=arXiv&color=FF0066&logo=arxiv" style="height: 25px;"></a>  
  <br>
  <a href="https://docs.google.com/presentation/d/1Wd2alQk565YBZjTaoVdSrdDacb_ILhlXTOzTTP_tTt4/edit?usp=sharing"><img src="https://img.shields.io/static/v1?label=Slides&message=Presentation&color=fa9f1b&logo=googleslides" style="height: 25px;"></a>   
  <a href="https://github.com/SKKUAutoLab/TSBOW"><img src="https://img.shields.io/static/v1?label=Code&message=Github&color=6699FF&logo=github" style="height: 25px;"></a>   
  <a href="https://huggingface.co/datasets/SKKUAutoLab/TSBOW"><img src="https://img.shields.io/static/v1?label=Dataset&message=HuggingFace&color=FF6600&logo=huggingface" style="height: 25px;"></a>  
</div>


![TSBOW Dataset Scenes](./Figure_All_Scenes.png)


<!-- MARK: News -->

## 🎉 NEWS

<!-- + [2025.12.31] 🔥 Our paper, code and TSBOW dataset are released! -->

+ [2026.07.15] 🧩 A camera_held_out subset is available on HuggingFace.
+ [2026.01.16] 📦 TSBOW dataset is available on HuggingFace.
+ [2025.11.16] 💻 Our code and website are released!
+ [2025.11.08] 🏆 **<span style="color: #FFCC00">T</span><span style="color: #33CCCC">S</span><span style="color: #FF6600">B</span><span style="color: #6699FF">O</span><span style="color: #FF0066">W</span>** has been accepted to **AAAI 2026**!



## 🎯 Associated Paper

<!-- **<span style="color: #FFCC00">T</span><span style="color: #33CCCC">S</span><span style="color: #FF6600">B</span><span style="color: #6699FF">O</span><span style="color: #FF0066">W</span>** Benchmark Datasets -->

This dataset is associated with the paper [TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions](https://ojs.aaai.org/index.php/AAAI/article/view/37439) (accepted in AAAI-26 Main Technical Track by The 40th Annual AAAI Conference on Artificial Intelligence).

<!-- For supplementary materials,  -->



<!-- MARK: Overview -->

## 🌍 Overview

**Comprehensive, annotated dataset for object detection.**
This dataset consists of over 32 hours of real-world traffic surveillance data across 71 CCTV and an additional color cameras, spanning annual weather conditions ([See Demo Videos](https://skkuautolab.github.io/TSBOW/TSBOW_scenes.html)). 
The UI for filtering scenes according to each attribute is provided in [TSBOW-Filter-Scenes](https://ngochdm.github.io/TSBOW-Scenes-Public/).

- *<span style="color: #FFCC00; font-weight: bold;">Hardware</span>*: CCTV system + color camera. 
- *<span style="color: #33CCCC; font-weight: bold;">Tasks</span>:* Object Detection.
- *<span style="color: #FF6600; font-weight: bold;">Position</span>:* South Korea. 
- *<span style="color: #6699FF; font-weight: bold;">Weather</span>:* sunny/cloudy, haze, rain, snow. 
- *<span style="color: #FF0066; font-weight: bold;">Daytime</span>:* day.


<!-- ## 📝 Abstract -->
<details>
    <summary>🖌️ Abstract</summary>
    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 <b>T</b>raffic <b>S</b>urveillance <b>B</b>enchmark for <b>O</b>ccluded Vehicles under Various <b>W</b>eather Conditions (<b>TSBOW</b>), a comprehensive dataset designed to enhance occluded vehicle detection across diverse annual weather scenarios. Comprising over <b>32 hours</b> of real-world traffic data from densely populated urban areas, TSBOW includes more than <b>48,000 manually annotated</b> and <b>3.2 million semi-labeled frames</b>; 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. <br>
    <b>Code</b> -- https://github.com/SKKUAutoLab/TSBOW
</details>


<details>
    <summary>📊 Statistics</summary>

<div align="center" style="max-width:1000px; margin: 10px auto 20px;">
    <div style="display:flex; gap:18px; justify-content:center; align-items:flex-start; flex-wrap:wrap;">
        <div style="flex:1 1 440px; max-width:51%; text-align:center;">
            <img src="./Figure_Suwon_Camera_Map.png" alt="Recording Locations" style="width:100%; height:auto; display:block; border-radius:6px;">
            <p style="margin:8px 0 0 0; font-weight:600;">Recording Locations</p>
        </div>
        <div style="flex:1 1 440px; max-width:42%; text-align:center;">
            <img src="./Chart_SunburstChart_Attributes.png" alt="Video Distribution" style="width:100%; height:auto; display:block; border-radius:6px;">
            <p style="margin:8px 0 0 0; font-weight:600;">Video Distribution</p>
        </div>
    </div>
</div>
</details>


<!-- Table Stats -->
<div align="center">
    <table style="width: 100%; max-width: 800px; margin: 30px auto; border-collapse: collapse;">
        <tr>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #33CCCC; font-weight: bold;">198</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #33CCCC; font-weight: bold;">🎞️ Processed Videos 🎞️</p>
            </td>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #FFCC00; font-weight: bold;">32 h</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #FFCC00; font-weight: bold;">⏱️ Duration ⏱️</p>
            </td>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #6699FF; font-weight: bold;">3.2 M</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #6699FF; font-weight: bold;">🖼️ Total Frames 🖼️</p>
            </td>
        </tr>
        <tr>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #FF6600; font-weight: bold;">71.1 M</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #FF6600; font-weight: bold;">Semi-Annotated<br>Instances 🤖</p>
            </td>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #33CCFF; font-weight: bold;">48 K</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #33CCFF; font-weight: bold;">Manual-Annotated<br>Frames 📝</p>
            </td>
            <td style="padding: 25px 15px; text-align: center; border: 1px solid #ddd; background: #092030;">
                <h2 style="margin: 0; font-size: 2.5em; color: #FF0066; font-weight: bold;">1.1 M</h2>
                <p style="margin: 8px 0 0 0; font-size: 1.1em; color: #FF0066; font-weight: bold;">Manual-Annotated<br>Instances 🎯</p>
            </td>
        </tr>
    </table>
</div>



<!-- MARK: Download -->

## 📥 Dataset Download

The **<span style="color: #FFCC00">T</span><span style="color: #33CCCC">S</span><span style="color: #FF6600">B</span><span style="color: #6699FF">O</span><span style="color: #FF0066">W</span>** dataset is distributed under the <a href="https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode.en"> Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International</a> License. 
By completing the form below, users acknowledge and agree that <b>the dataset will be used solely for research purposes</b>. 

<details>
    <summary><b>Submission Guidelines</b></summary>

<b>1. TSBOW - Terms and Conditions Form</b>
- Download and fill out the <a href="https://docs.google.com/document/d/1Mn4qV8HErCX1EMQvys6SAKbMb06dZa85nie7rzSxeGc/edit?usp=sharing"> TSBOW - Terms and Conditions</a> form. 
- Ensure all **User Information** fields are completed.
- Provide a description of your intended use of the dataset.
- Check the agreement box and insert your handwritten signature.
- Save as PDF file. Renaming as: **{TSBOW}_{Application-Date}_{Huggingface-Username}.pdf** (i.e. TSBOW_20260120_ngochdm.pdf)

<b>2. Requirement Email</b>
- The subject format: [TSBOW Access Requirement] {Your Name} - {Affiliation}
- The email body includes your <b>HuggingFace account information</b> (username and email). We will verify this information against the access requirements on Hugging Face before approval.
- Send email to all our addresses: <b>jwjeon@skku.edu, ngochdm@skku.edu, automation.skku@gmail.com</b>

<b>3. Send Request on HuggingFace</b>
- Press "Agree and send request to access TSBOW" button <a href="https://huggingface.co/datasets/SKKUAutoLab/TSBOW">on HuggingFace</a>. Our team may take 2-3 days to process your request.

</details>

Scripts to download **<span style="color: #FFCC00">T</span><span style="color: #33CCCC">S</span><span style="color: #FF6600">B</span><span style="color: #6699FF">O</span><span style="color: #FF0066">W</span>** from HuggingFace are provided in [[Github] utils](https://github.com/SKKUAutoLab/TSBOW/tree/main/utils) folder. Please refer to the [`download_TSBOW.py`](https://github.com/SKKUAutoLab/TSBOW/tree/main/utils/download_TSBOW.py) for more details.



<!-- MARK: Format -->

## 📁 Format

### Video Format

- Video Standard: mp4 (H.265)
- Video Resolution: 1280 x 720
- Video Frame Rate: varied from 20 to 30 FPS
- Videos are named as `<video_id>.mp4`


### Image Format

- Image Standard: jpg
- Image Resolution: 1280 x 720
- Images extracted from videos are named as `<video_id>_<frame_id>.jpg`. `<frame_id>` is 6 characters length.


### Ground Truth Format

Annotations are provided in YOLO format with one *.txt file per image. If there are no objects in an image, no *.txt file is required.
Each line in text file are the bounding box information of an object:

```
<class_id> <x_center> <y_center> <width> <height>
```

where
- `class_id` (int): class index starting from 0
- `x_center`, `y_center` (double): box coordinates normalized xywh format (from 0 to 1)
- `width`, `height` (double): box width, height


### Metadata Format

<!-- TSBOW_info.csv -->
The [CSV file](metadata/TSBOW_info.csv) provides metadata for each videos, including: scenario, daytime, weather, scale, roadtype, video_id, and total_duration. 

- `SCENARIO` (char): one of four values: `r` (road), `i` (intersection), `s` (special cases), `d` (disaster).
- `DAYTIME` (char): `d` (day), `n` (night).
- `WEATHER` (char): one of four values: `n` (normal), `h` (haze), `r` (rain), `s` (snow).
- `SCALE` (char): one of three values: `f` (fine), `m` (medium), `c` (coarse).
- `ROADTYPE` (char): one of three values: `u` (urban), `s` (standard), `b` (boulevard).
- `VIDEO_ID` (string): video name.
- `DURATION` (duration): total duration of whole videos (test + val + train). 


<!-- TSBOW.yaml -->
The [TSBOW.yaml](v1.0.0_official_release/TSBOW.yaml) provides information for training models on training and validation set. 

- `path` (string): path to dataset.
- `train`, `val`, `test` (string): name of txt file containing list of train/val/test images. These files are also provided in **<release_version>** folder.
- `names`: a list of object categories and their indexes. The order of class list is according to the order of class name in `class.txt`.



### Directory Structure

- **Train/Val/Test_Public**: 
    - `videos.zip`: video files. 
    - `annotations.zip`: annotations include images and labels.
    - `semilabels.zip`: semi-labeled annotations include labels only. You can extract frames from videos.
    - Notes for Test_Public: In the first publication, we only public annotations of the last 30% of test set. The remaining files will be public in the future.

- **Others**:
    - `comparison.zip`: annotations of the comparison subset in Experiment section, include images and labels.
    - `TSBOW_info.csv`: metadata file.
    - `classes.txt`: a list of annotated object categories. Class definition is mentioned in [TSBOW_Class_Definition](https://github.com/SKKUAutoLab/TSBOW/blob/main/documents/TSBOW_Class_Definition.pdf).



<!-- MARK: Citation -->

## 🏅 Citation

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

```bibtex
@article{Huynh2026TSBOW, 
    title={TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions}, 
    author={Huynh, Ngoc Doan-Minh and Tran, Duong Nguyen-Ngoc and Pham, Long Hoang and Tran, Tai Huu-Phuong and Jeon, Hyung-Joon and Nguyen, Huy-Hung and Khac Vu, Duong and Jeon, Hyung-Min and Phan, Son Hong and Pham-Nam Ho, Quoc and Tran, Chi Dai and Khanh, Trinh Le Ba and Jeon, Jae Wook}, 
    journal={Proceedings of the AAAI Conference on Artificial Intelligence}, 
    volume={40}, 
    number={7}, 
    url={https://ojs.aaai.org/index.php/AAAI/article/view/37439}, 
    DOI={10.1609/aaai.v40i7.37439}, 
    year={2026}, 
    month={Mar.}, 
    pages={5239-5247} 
}
```

<!-- ```bibtex
@misc{huynh2026tsbowtrafficsurveillancebenchmark,
      title={TSBOW: Traffic Surveillance Benchmark for Occluded Vehicles Under Various Weather Conditions}, 
      author={Ngoc Doan-Minh Huynh and Duong Nguyen-Ngoc Tran and Long Hoang Pham and Tai Huu-Phuong Tran and Hyung-Joon Jeon and Huy-Hung Nguyen and Duong Khac Vu and Hyung-Min Jeon and Son Hong Phan and Quoc Pham-Nam Ho and Chi Dai Tran and Trinh Le Ba Khanh and Jae Wook Jeon},
      year={2026},
      eprint={2602.05414},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2602.05414}, 
}
``` -->



## 📝 Changelog

- Jul 15, 2026: Release subset: camera_held_out (v1.1.0).
- Jul 08, 2026: Re-organize metadata folder.
- Jul 01, 2026: Re-organize dataset for the official_release version, named as v1.0.0_official_release.


<div align="center"><a href="#top">🔝 Back to Top</a></div>



<!-- HuggingFace Documents -->
<!-- https://huggingface.co/docs/hub/datasets-cards -->
<!-- https://huggingface.co/docs/hub/paper-pages -->
<!-- https://huggingface.co/docs/hub/en/models-download-stats -->
<!-- https://huggingface.co/docs/hub/datasets-gated -->



<!-- 
configs:
- config_name: train
  data_files:
    - train/videos.zip
    - train/annotations.zip
    - train/semilabels.zip
    - metadata/train.txt
  default: true
- config_name: val
  data_files:
    - val/videos.zip
    - val/annotations.zip
    - val/semilabels.zip
    - metadata/val.txt
- config_name: test_public
  data_files:
    - test_public/annotations.zip
    - metadata/test_public.txt
- config_name: metadata
  data_files: metadata/TSBOW_info.csv
  sep: ','
  default: true
- config_name: additional_data
  data_files: comparison.zip 
  
-->