UEMM-Air
收藏资源简介:
UEMM-Air是由河海大学开发的多模态无人机对象检测数据集,专注于无人机视角下的复杂背景和小对象检测。该数据集通过虚幻引擎模拟多种飞行场景和对象类型,自动收集不同场景、视角和高度的数据,并采用新颖的启发式自动标注算法生成精确的对象检测标签。UEMM-Air包含5种模态,共计20000对图像,适用于城市监控和军事侦察等领域的模型训练和评估。
UEMM-Air is a multimodal unmanned aerial vehicle (UAV) object detection dataset developed by Hohai University, focusing on complex background and small object detection from UAV perspectives. This dataset simulates various flight scenarios and object types using Unreal Engine, automatically collects data across different scenarios, viewpoints and altitudes, and generates accurate object detection labels with a novel heuristic automatic annotation algorithm. UEMM-Air includes 5 modalities and a total of 20,000 image pairs, which is suitable for model training and evaluation in fields such as urban surveillance and military reconnaissance.




