CARLA Simulator-based Synthetic Evaluation Dataset

本研究基于CARLA模拟器创建了一个新型数据集，用于测试在不同降水条件下多种网络模型的性能。该数据集包含从10mm/h到100mm/h的降水情况，共计2200张图像，用于评估YOLO系列模型在极端重雨条件下的物体检测性能。数据集的创建涉及使用实验获得的雨滴数据，并通过模拟合成雨天图像。此数据集主要应用于自动驾驶领域，旨在解决恶劣天气条件下物体检测性能下降的问题。

This study develops a novel dataset based on the CARLA simulator to test the performance of multiple network models under varying precipitation conditions. This dataset covers precipitation levels ranging from 10 mm/h to 100 mm/h, with a total of 2200 images, and is designed to evaluate the object detection performance of YOLO-series models under extreme heavy rainfall conditions. The creation of this dataset involves using experimentally obtained raindrop data and synthesizing rainy-day images through simulation. This dataset is primarily applied in the autonomous driving domain, aiming to address the issue of degraded object detection performance under adverse weather conditions.