EVD4UAV
收藏资源简介:
EVD4UAV是由克利夫兰州立大学创建的一个针对无人机图像中车辆检测的敏感性基准数据集。该数据集包含6284张图像和90,886个精细标注的车辆实例,涵盖了50m、70m和90m三种不同高度,以及车辆的颜色和类型等属性。数据集通过精细的标注,包括水平和旋转边界框以及实例级掩码,以顶视图形式呈现,确保车辆顶部清晰可见。EVD4UAV数据集旨在研究基于对抗性补丁的车辆检测攻击问题,特别是在不同高度下的攻击效果,为无人机图像中的车辆检测提供了一个全面的测试平台。
EVD4UAV is a sensitivity benchmark dataset for vehicle detection in unmanned aerial vehicle (UAV) imagery, created by Cleveland State University. This dataset comprises 6,284 images and 90,886 finely annotated vehicle instances, covering three distinct altitude levels (50m, 70m, and 90m) along with attributes such as vehicle color and type. All images in the dataset are presented in top-down view to ensure clear visibility of vehicle rooftops, with fine-grained annotations including horizontal and rotated bounding boxes as well as instance-level masks. The EVD4UAV dataset is designed to study adversarial patch-based vehicle detection attacks, particularly their attack performance under varying altitudes, providing a comprehensive testbed for vehicle detection in UAV imagery.

- 1EVD4UAV: An Altitude-Sensitive Benchmark to Evade Vehicle Detection in UAV克利夫兰州立大学 · 2024年



