无人机图像语义分割数据集

UAVid是面向无人机图像语义分割的高分辨率数据集，包含30个视频序列共300张4K图像，每帧均标注8个语义类别。数据以倾斜视角拍摄，具有大规模场景变化、运动物体丰富和时间连续性强等特点。适用于语义分割、动态场景理解与时间一致性建模，广泛应用于无人机测绘、建筑检测等实际场景。

UAVid is a high-resolution dataset for unmanned aerial vehicle (UAV) image semantic segmentation. It contains 30 video sequences totaling 300 4K images, with each frame annotated with 8 semantic categories. The dataset is captured from an oblique perspective, featuring large-scale scene variations, abundant moving objects and strong temporal continuity. It is applicable to semantic segmentation, dynamic scene understanding and temporal consistency modeling, and has been widely used in practical scenarios such as UAV mapping and building detection.

搜集汇总

背景与挑战

背景概述

UAVid是一个高分辨率的无人机语义分割数据集，包含30个视频序列的300张4K图像，每张图像均标注8个语义类别。数据以倾斜视角拍摄，具有大规模场景变化、运动物体丰富和时间连续性强等特点，适用于语义分割、动态场景理解和时间一致性建模等任务。该数据集由荷兰特文特大学、武汉大学和俄亥俄州立大学联合发布，总容量为6.70GB。