LanEvil
收藏资源简介:
LanEvil数据集由北京航空航天大学创建,专注于评估车道检测系统在面对环境幻觉时的鲁棒性。该数据集包含90,292张通过CARLA模拟器生成的图像,涵盖了14种关键的环境幻觉类型,如阴影和反射。创建过程中,研究团队精心设计了3D交通环境,并通过模拟自然环境变化来生成图像。LanEvil数据集的应用领域主要集中在自动驾驶系统中,旨在提高系统在复杂环境下的识别准确性和安全性,解决因环境幻觉导致的误识别问题。
The LanEvil dataset, developed by Beihang University, is dedicated to evaluating the robustness of lane detection systems against environmental hallucinations. It contains 90,292 images generated through the CARLA simulator, covering 14 key types of environmental hallucinations including shadows and reflections. During the creation process, the research team meticulously designed a 3D traffic environment and generated images by simulating natural environmental changes. The primary application scenario of the LanEvil dataset is autonomous driving systems, aiming to enhance the recognition accuracy and safety of such systems in complex environments and mitigate misidentification issues caused by environmental hallucinations.




