Amodal Cityscapes

Amodal Cityscapes数据集由布伦瑞克工业大学通信技术研究所创建，专注于自动驾驶中的模态感知任务，特别是解决遮挡问题。该数据集基于Cityscapes数据集，包含5000张城市交通场景图像，用于训练和验证模态语义分割方法。数据集通过复制粘贴实例到目标图像中来合成，以模拟模态标签，旨在提供像素级的遮挡信息。该数据集的应用领域主要集中在自动驾驶的环境感知，旨在提高车辆在复杂城市环境中的感知能力，解决遮挡带来的挑战。

The Amodal Cityscapes dataset was developed by the Institute of Communication Technology, Technische Universität Braunschweig. It focuses on the amodal perception task in autonomous driving, specifically addressing the occlusion challenge. Built upon the original Cityscapes dataset, this dataset includes 5000 urban traffic scene images for training and validating amodal semantic segmentation methods. The dataset is synthesized via copy-pasting object instances into target images to generate amodal labels, with the aim of providing pixel-level occlusion information. Its application domains are primarily centered around environmental perception for autonomous driving, with the goal of improving vehicles' perception capabilities in complex urban environments and overcoming challenges caused by occlusions.