![[2602.12902] Robustness of Object Detection of Autonomous Vehicles in Adverse Weather Conditions](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.12902] Robustness of Object Detection of Autonomous Vehicles in Adverse Weather Conditions
Summary
This paper evaluates the robustness of object detection models used in autonomous vehicles under adverse weather conditions, proposing a method to assess model performance through synthetic data generation.
Why It Matters
As autonomous vehicles become more prevalent, ensuring their safety in various environmental conditions is crucial. This research provides insights into how different object detection models perform under adverse weather, which is vital for public safety and technology reliability.
Key Takeaways
- The study introduces a method for evaluating object detection robustness in adverse weather.
- Faster R-CNN outperformed other models with an average first failure coefficient of 71.9%.
- Synthetic data training can enhance model robustness but may lead to diminishing returns if overused.
- The research highlights the importance of assessing models under realistic adverse conditions.
- Data augmentation techniques were effectively used to simulate various weather scenarios.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.12902 (cs) [Submitted on 13 Feb 2026] Title:Robustness of Object Detection of Autonomous Vehicles in Adverse Weather Conditions Authors:Fox Pettersen, Hong Zhu View a PDF of the paper titled Robustness of Object Detection of Autonomous Vehicles in Adverse Weather Conditions, by Fox Pettersen and Hong Zhu View PDF HTML (experimental) Abstract:As self-driving technology advances toward widespread adoption, determining safe operational thresholds across varying environmental conditions becomes critical for public safety. This paper proposes a method for evaluating the robustness of object detection ML models in autonomous vehicles under adverse weather conditions. It employs data augmentation operators to generate synthetic data that simulates different severance degrees of the adverse operation conditions at progressive intensity levels to find the lowest intensity of the adverse conditions at which the object detection model fails. The robustness of the object detection model is measured by the average first failure coefficients (AFFC) over the input images in the benchmark. The paper reports an experiment with four object detection models: YOLOv5s, YOLOv11s, Faster R-CNN, and Detectron2, utilising seven data augmentation operators that simulate weather conditions fog, rain, and snow, and lighting conditions of dark, bright, flaring, and shadow. The experiment data show that the method is feasible, effect...
