DDOS
收藏资源简介:
DDOS数据集,由帝国理工学院开发,专注于无人机视角下的深度和障碍物分割任务。该数据集通过高级计算机图形技术生成了34000个合成图像,特别强调对细小物体的检测和分割,如电线和栅栏。创建过程中,数据集利用了精确的像素级标注,包括深度图、光学流和表面法线,以支持精确的算法训练和评估。DDOS数据集的应用领域主要集中在提升无人机导航的安全性和效率,特别是在识别和规避飞行路径中的细小障碍物方面。
The DDOS Dataset, developed by Imperial College London, focuses on depth and obstacle segmentation tasks from unmanned aerial vehicle (UAV) perspectives. It generates 34,000 synthetic images via advanced computer graphics techniques, with particular emphasis on the detection and segmentation of small objects such as wires and fences. During its development, the dataset leverages precise pixel-level annotations including depth maps, optical flow, and surface normals to support rigorous algorithm training and evaluation. The primary application domains of the DDOS Dataset aim to enhance the safety and efficiency of UAV navigation, specifically in identifying and evading tiny obstacles along flight paths.

- 1DDOS: The Drone Depth and Obstacle Segmentation Dataset帝国理工学院 · 2023年



