The first open-source synthetic dataset for collaborative perception focused on adverse weather conditions.
Our paper "Adver-City: Open-Source Multi-Modal Dataset for Collaborative Perception Under Adverse Weather Conditions" is available on Arxiv.
Adverse weather conditions pose a significant challenge to the widespread adoption of Autonomous Vehicles (AVs) by impacting sensors like LiDARs and cameras. Even though Collaborative Perception (CP) improves AV perception in difficult conditions, existing CP datasets lack adverse weather conditions. To address this, we introduce Adver-City, the first open-source synthetic CP dataset focused on adverse weather conditions. Simulated in CARLA with OpenCDA, it contains over 24 thousand frames, over 890 thousand annotations, and 110 unique scenarios across six different weather conditions: clear weather, soft rain, heavy rain, fog, foggy heavy rain and, for the first time in a synthetic CP dataset, glare. It has six object categories including pedestrians and cyclists, and uses data from vehicles and roadside units featuring LiDARs, RGB and semantic segmentation cameras, GNSS, and IMUs. Its scenarios, based on real crash reports, depict the most relevant road configurations for adverse weather and poor visibility conditions, varying in object density, with both dense and sparse scenes, allowing for novel testing conditions of CP models. Benchmarks run on the dataset show that weather conditions created challenging conditions for perception models, reducing multi-modal object detection performance by up to 19%, while object density affected LiDAR-based detection by up to 29%. The dataset, code and documentation are available at https://labs.cs.queensu.ca/quarrg/datasets/adver-city/.
The Adver-City dataset may be downloaded from its official website. Check the Data structure documentation for details on how the dataset files are organized.
We follow an annotation schema compatible with OPV2V. For detailed information, check the Annotation documentation.
Adver-City was built using CARLA v0.9.12 and OpenCDA, so check OpenCDA's installation tutorial for detailed instructions on how to install CARLA and the necessary packages.
Besides the packages used by OpenCDA, Adver-City also Pandas 1.3.5, which can be installed by running:
pip install pandas==1.3.5Pandas is only required for generating statistics of the scenarios.
To run all scenarios, generating their videos (-v) and summaries (-m), run:
python main.py -v y -m yHaving the summary files for all scenarios, the statistics may be generated by running the following command with the
appropriate path argument (-p):
python generate_statistics.py -p data_dumping/2024_06_14_12_47_41To run a single scenario, select a road configuration (-s), weather condition (-w) and density setting (-d) and
run the command with their abbreviations (refer to the Scenarios section below). For the Urban Intersection Clear Night
Sparse scenario, the command would be:
python main.py -s ui -w cn -d s -v y -m yAdver-City's scenarios provide a rich testbed for comparing how models perform on varying environmental conditions. We combine 5 road configurations, 11 weather and daytime conditions and 2 density settings to create a total of 110 unique scenarios.
| Road Configuration | Abbreviation |
|---|---|
| Urban Intersection | ui |
| Urban Non-Junction | unj |
| Rural Intersection | ri |
| Rural Straight Non-Junction | rsnj |
| Rural Curved Non-Junction | rcnj |
| Weather and Daytime Condition |
Abbreviation |
|---|---|
| Clear Day | cd |
| Clear Night | cn |
| Soft Rain Day | srd |
| Soft Rain Night | srn |
| Heavy Rain Day | hrd |
| Heavy Rain Night | hrn |
| Foggy Day | fd |
| Foggy Night | fn |
| Foggy Heavy Rain Day | fhrd |
| Foggy Heavy Rain Night | fhrn |
| Glare Day | gd |
| Density Setting | Abbreviation |
|---|---|
| Dense | d |
| Sparse | s |
For minor modifications, check the Scenarios documentation.
For more in-depth modifications, check our documentation for a description on how Adver-City's scenario generation code works.
Feel free to fork or create pull requests with your modifications!
If this work is helpful for your research, please cite the following BibTeX entry.
@article{Karvat_2024_AdverCity,
title = {Adver-City: Open-Source Multi-Modal Dataset for Collaborative Perception Under Adverse Weather Conditions},
author = {Karvat, Mateus and Givigi, Sidney},
journal = {arXiv preprint arXiv:2410.06380},
note = {Available at https://arxiv.org/abs/2410.06380},
year = {2024}
}