Spot-the-diff

该数据集使用了真实世界的图像对，重点关注时间演变以生成修改后的图像。它倾向于包含停车场内的图片，这使得地面真实变化类型的多样性受到了限制。该数据集的任务是进行图像差异描述。

This dataset utilizes real-world image pairs, with a primary focus on temporal evolution for generating modified images. It predominantly contains images captured within parking lots, which limits the diversity of ground-truth change types. The core task of this dataset is image difference description.