AI-TOD-v2
收藏资源简介:
AI-TOD-v2数据集是由武汉大学电子信息学院创建的,专门用于航空图像中的微小物体检测。该数据集包含28,036张图像和752,754个实例,平均物体大小仅为12.7像素,是所有目标检测数据集中物体尺寸最小的。AI-TOD-v2数据集通过精心重新标注,显著减少了缺失标注和位置错误的问题,提高了训练和验证过程的可靠性。该数据集的应用领域包括车辆检测、交通状况监控和海上救援等,旨在解决航空图像中微小物体检测的挑战。
The AI-TOD-v2 dataset was constructed by the School of Electronic Information, Wuhan University, and is dedicated to tiny object detection in aerial imagery. Comprising 28,036 images and 752,754 instances, it boasts an average object size of merely 12.7 pixels, making it the object detection dataset with the smallest average object scale among all existing datasets. Through meticulous re-annotation, the AI-TOD-v2 dataset substantially mitigates issues of missing annotations and positional inaccuracies, thus enhancing the reliability of its training and validation procedures. Its applicable scenarios cover vehicle detection, traffic state monitoring, maritime rescue, and other related fields, aiming to address the core challenges of tiny object detection in aerial images.




