microsoft/IMAGE_UNDERSTANDING

这是一个用于测试空间推理、视觉提示以及对象识别和检测能力的程序生成合成数据集。数据集包含四个子任务：对象识别、视觉提示、空间推理和对象检测，每种任务都有单一和成对两种条件，适用于含有单个或两个对象的图像。数据集中的图像是由COCO对象列表中的对象粘贴在随机的Places365背景图像上构成的，每个对象在图像中的位置都有随机性，并伴有轻微的旋转、位置抖动和缩放。

This is a procedurally generated synthetic dataset designed to test the abilities of spatial reasoning, visual prompting, as well as object recognition and detection. The dataset consists of four sub-tasks: Object Recognition, Visual Prompting, Spatial Reasoning, and Object Detection, each with single and pair conditions for images containing one or two objects. The images in the dataset are composed of objects from the COCO object list pasted onto random Places365 background images, with each objects position in the image being random and accompanied by slight rotation, positional jitter, and scaling.