AI-TOD

该数据集名为AI-TOD，包含了用于目标检测的航空影像，涵盖了8种类别的目标，共计700,621个对象实例。与其他数据集相比，AI-TOD图像的平均尺寸为12.8像素，相对较小。该数据集规模为28,036张航空影像，其中14,018张作为训练集，另外14,018张作为测试集，其任务是进行目标检测。

The dataset named AI-TOD contains aerial images for object detection, covering 8 object categories with a total of 700,621 object instances. Compared with other datasets, the average size of images in AI-TOD is 12.8 pixels, which is relatively small. The dataset consists of 28,036 aerial images in total, with 14,018 used as the training set and another 14,018 as the test set, and its task is object detection.

搜集汇总

背景与挑战

背景概述

AI-TOD是一个专注于航空图像中微小物体检测的数据集，包含700,621个对象实例和28,036张图像，平均对象大小仅为12.8像素。数据集需要结合xView训练集和AI-TOD_wo_xview部分，通过专用工具生成完整数据。