mair-lab/escher-vismin-dev

Vismin数据集是一个单语言（英文）的数据集，包含原始图像和编辑后的图像，以及编辑指令和关于编辑的详细、简洁和非常简洁的推理描述。数据集的任务类别是图像到图像的任务，具体任务为图像修复。数据集大小小于1K，且数据集是原创的，许可证为MIT。数据实例还包括源图像和目标图像的标题以及额外的元数据字段。The Vismin dataset is a monolingual (English) dataset containing original and edited images, along with edit instructions and detailed, concise, and super concise reasoning descriptions of the edit. The datasets task category is image-to-image tasks, specifically image inpainting. The dataset size is less than 1K, and it is original with an MIT license. Data instances also include captions for the source and target images, as well as additional metadata fields.