SVGEditBench V2
收藏资源简介:
SVGEditBench V2是一个用于指令基础SVG编辑的基准数据集,由东京大学创建。该数据集包含原始SVG图像、编辑后的地面真实SVG图像和编辑提示组成的 triplet。通过从多个公开的SVG表情符号数据集中提取图像对,并使用GPT-4o生成编辑提示构建而成。数据集涵盖了从简单的颜色更改到复杂的元素转换等多种编辑任务,适用于评估SVG编辑方法的效果。
SVGEditBench V2 is a benchmark dataset for instruction-based SVG editing, developed by The University of Tokyo. This dataset consists of triplets composed of original SVG images, edited ground-truth SVG images and corresponding editing prompts. It is constructed by extracting image pairs from multiple publicly available SVG emoji datasets and generating editing prompts using GPT-4o. The dataset covers a wide range of editing tasks, from simple color modifications to complex element transformations, and is suitable for evaluating the performance of SVG editing methods.




