WildDash

我们包含来自世界各地的各种数据源的图像，这些图像具有许多不同的困难场景（例如下雨、道路覆盖、黑暗、过度曝光）和相机特征（噪声、压缩伪影、失真）。提供的地面实况格式与 Cityscapes 兼容。 WildDash 2 还引入了对新标签的支持，例如：货车、皮卡、路灯、广告牌、道路标记和自我车辆。
该数据集的主要重点是测试。它包含在真实世界驾驶情况下记录的数据。它的目标是：
编制并提供可用于评估的标准数据。
建立公认的评估协议、数据和措施。
使用计算机视觉技术促进驾驶应用程序的算法开发。

This dataset comprises images from diverse data sources worldwide, featuring numerous challenging scenarios (e.g., rainy conditions, road coverage, low light, overexposure) and camera-related characteristics (noise, compression artifacts, distortion). The provided ground truth format is compatible with Cityscapes. WildDash 2 additionally introduces support for new annotation labels, including vans, pickup trucks, streetlights, billboards, road markings, and ego vehicles.

The core focus of this dataset is testing. It contains data recorded in real-world driving scenarios. Its intended goals are as follows:
1. Compile and provide standard data suitable for evaluation tasks;
2. Establish widely accepted evaluation protocols, datasets, and metrics;
3. Facilitate the development of algorithms for driving applications using computer vision technologies.