| --- |
| license: mit |
| task_categories: |
| - feature-extraction |
| language: |
| - en |
| tags: |
| - perception |
| - cooperative |
| - collective |
| - autonomous_driving |
| pretty_name: >- |
| CoopScenes: Multi-Scene Infrastructure and Vehicle Data for Advancing |
| Collective Perception in Autonomous Driving |
| viewer: false |
| --- |
| |
|
|
| # 📚 CoopScenes Dataset |
|
|
|  |
|
|
| ## 🌟 Overview |
|
|
| **CoopScenes** is a comprehensive multi-scene dataset designed for research in collective perception, sensor registration, and cooperative systems in urban environments. It features synchronized data from an ego vehicle and infrastructure-mounted sensors across real-world scenarios, including public transport stops, construction sites, and expressways. |
|
|
| - **Duration**: \~104 minutes at 10 Hz → \~62,000 frames (\~527 GB in `.4mse` format) |
| - **Synchronization**: Sub-frame alignment with \~2.3 ms mean offset |
| - **Scenarios**: Collected across multiple cities in the Stuttgart metropolitan area |
|
|
| 🔜 More information: [coopscenes.github.io](https://coopscenes.github.io/) |
|
|
| --- |
|
|
| ## 🛠️ Sensor Setup & Annotations |
|
|
| The dataset features time-synchronized and spatially calibrated sensors on both the ego vehicle and roadside infrastructure (towers), including: |
|
|
| - LiDAR (Ouster OS2, Blickfeld Qb2) |
| - Multi-camera systems |
| - GNSS and IMU |
| - Object annotations (automatically generated) |
| - Privacy-preserving anonymization using [**BlurScene**](https://pypi.org/project/BlurScene/) |
|
|
| --- |
|
|
| ## ✅ Key Features |
|
|
| | Feature | Description | |
| | ------------------------------------- | ----------------------------------------------- | |
| | 62,000 Frames at 10 Hz | \~104 minutes of data | |
| | High-precision synchronization | Mean offset \~2.3 ms | |
| | Vehicle-to-infrastructure setup | Multi-agent cooperative perception | |
| | Diverse scenarios | Public transport, construction, highways | |
| | Automatic annotations & anonymization | Faces and license plates blurred with BlurScene | |
|
|
| --- |
|
|
| ## 📦 Installation & Usage |
|
|
| Install the CoopScenes Python package: |
|
|
| ```bash |
| pip install coopscenes |
| ``` |
|
|
| Then load and explore the dataset using the included developer tools: |
|
|
| ```python |
| from coopscenes import DataRecord |
| |
| # open a specific .4mse-file |
| record = DataRecord("/content/example_record_1.4mse") |
| # use first frame from record |
| frame = record[0] |
| |
| frame.vehicle.cameras.STEREO_LEFT.show() |
| print(frame.tower.lidars.UPPER_PLATFORM.points.shape) # example access |
| ``` |
|
|
| Additional tooling, documentation, and format specs can be found in the [developer toolkit](https://pypi.org/project/coopscenes/). |
|
|
| --- |
|
|
| ## 🚀 Google Colab (Quickstart) |
|
|
| Get started with the data using our ready-to-run [**Colab notebook**](https://coopscenes.github.io/#colab). It demonstrates: |
|
|
| - Reading `.4mse` files |
| - Visualizing sensor data |
| - Performing simple analysis tasks |
|
|
| --- |
|
|
| ## 📄 Citation |
|
|
| Please cite the following if you use CoopScenes in your work (IEEE IV '25 publish is following): |
|
|
| ```bibtex |
| @misc{vosshans2025aeifdatacollectiondataset, |
| author = {Marcel Vosshans and Alexander Baumann and Matthias Drueppel and Omar Ait-Aider and Youcef Mezouar and Thao Dang and Markus Enzweiler}, |
| title = {CoopScenes: Multi-Scene Infrastructure and Vehicle Data for Advancing Collective Perception in Autonomous Driving}, |
| url = {https://arxiv.org/abs/2407.08261}, |
| year = {2025}, |
| } |
| ``` |
|
|
| --- |
|
|
| ## 🔐 License |
|
|
| The dataset is released under the **MIT License**. Refer to the LICENSE file for details. |
|
|
|
|
