UAVVaste

该数据集名为VisDrone-2019-DET，包含了由无人机拍摄的图像，并对十个物体类别进行了边界框标注，旨在促进对小物体的检测。该数据集涵盖了行人、人、汽车、货车、公交车、卡车、摩托车、自行车、带篷三轮车和三轮车等十类物体的注释。规模上，它包含了6471张训练图像、548张验证图像以及3190张测试图像。该数据集的任务是进行物体检测。

The dataset named VisDrone-2019-DET comprises images captured by unmanned aerial vehicles (UAVs), with bounding box annotations for ten object categories, and is designed to facilitate research on small object detection. It provides annotations for ten types of objects: pedestrians, people, cars, vans, buses, trucks, motorcycles, bicycles, tricycles with canopies, and tricycles. In terms of scale, it contains 6,471 training images, 548 validation images, and 3,190 test images. The core task of this dataset is object detection.

搜集汇总

背景与挑战

背景概述

UAVVaste数据集用于评估小目标检测模型的性能，特别是SO-DETR系列模型在该数据集上的表现。实验结果显示，不同版本的SO-DETR模型在UAVVaste数据集上均取得了较高的AP和AP50值，其中SO-DETR-R50表现最佳，AP为37.5，AP50为76.4。