WEDGE
收藏资源简介:
WEDGE数据集是由卡内基梅隆大学的研究团队基于DALL-E生成模型创建的,旨在提升自动驾驶系统在极端天气条件下的感知能力。该数据集包含3360张合成图像,覆盖16种不同的极端天气场景,如雪、雨、雾等,每种天气类别包含210张图像。数据集通过精心设计的提示(prompts)生成,确保图像内容与天气条件紧密相关,并提供了详细的2D边界框标注,支持物体检测和天气分类任务。WEDGE数据集的应用领域主要集中在提升自动驾驶车辆在复杂天气环境下的感知和决策能力,解决现有数据集在极端天气条件下表现不足的问题。
The WEDGE dataset was developed by a research team from Carnegie Mellon University using the DALL-E generative model, with the goal of enhancing the perception capabilities of autonomous driving systems under extreme weather conditions. This dataset comprises 3,360 synthetic images covering 16 distinct extreme weather scenarios including snow, rain, fog, and others, with 210 images for each weather category. Generated through carefully crafted prompts, the dataset ensures that the image content is closely tied to the corresponding weather conditions, and is equipped with detailed 2D bounding box annotations to support object detection and weather classification tasks. The primary application scope of the WEDGE dataset lies in improving the perception and decision-making abilities of autonomous vehicles in complex weather environments, addressing the limitations of existing datasets when operating under extreme weather conditions.

- 1WEDGE: A multi-weather autonomous driving dataset built from generative vision-language models卡内基梅隆大学 · 2023年



