UAVid-RIS
收藏资源简介:
UAVid-RIS数据集是一个用于无人机基础上指向图像分割的新颖数据集,从原始的UAVid数据集派生而来,专为在复杂城市环境中进行无人机场景理解设计。该数据集包含了大约45度角的斜视航空影像,这些影像提供了比俯视图像更丰富的上下文信息。数据是从无人机在大约50米高空的4K视频记录中提取的高分辨率帧。影像内容覆盖了多种街道场景,包括车辆、行人、建筑物、道路、植被、广告牌和交通基础设施等对象。所有图像都被裁剪成1024×1024像素的大小,并且在生成过程中应用了类别特定的最小面积阈值以确保有意义的补丁选择。由于处理后的数据集无法重新分发,用户需要从UAVid官方网站下载原始数据集,并按照AeroReformer提供的预处理步骤生成所需数据。
The UAVid-RIS dataset is a novel dataset for UAV-based oblique image segmentation, derived from the original UAVid dataset and specifically designed for UAV scene understanding in complex urban environments. It contains oblique aerial imagery captured at an approximate 45-degree angle, which provides richer contextual information than nadir-looking top-down images. The data consists of high-resolution frames extracted from 4K video recordings taken by UAVs flying at an altitude of around 50 meters. The imagery covers diverse street scenes, including objects such as vehicles, pedestrians, buildings, roads, vegetation, billboards, and traffic infrastructure. All images are cropped to a size of 1024×1024 pixels, and class-specific minimum area thresholds are applied during the dataset generation process to ensure the selection of meaningful patches. Since the processed dataset cannot be redistributed, users need to download the original dataset from the official UAVid website and generate the required data following the preprocessing steps provided by AeroReformer.




