151万组单图&多图融合图像编辑数据

151万组单图&多图融合图像编辑数据。编辑类型包括人像/物体一致性编辑50万组、结构性编辑30万组、混合编辑21万组、空间编辑45万组，风格迁移5万组。编辑目标涵盖人物、动物、商品、植物、风景等场景。在标注方面，根据编辑指令，对图像中需要编辑的目标进行编辑。数据可用于图像合成、数据增强、虚拟场景生成等任务。

This dataset contains 1,510,000 groups of single-image and multi-image fusion image editing data. The editing types include 500,000 groups of human/object consistency editing, 300,000 groups of structural editing, 210,000 groups of hybrid editing, 450,000 groups of spatial editing, and 50,000 groups of style transfer. The editing targets cover scenarios such as humans, animals, commodities, plants, landscapes and others. Regarding annotation, the targets requiring editing in the images are labeled according to the provided editing instructions. This dataset can be used for tasks including image synthesis, data augmentation, virtual scene generation and more.

搜集汇总

背景与挑战

背景概述

该数据集包含151万组单图和多图融合的图像编辑数据，涵盖人像/物体一致性编辑、结构性编辑、混合编辑、空间编辑和风格迁移等多种编辑类型，目标场景包括人物、动物、商品、植物和风景等。数据经过像素级标注，准确率高，适用于图像合成、数据增强和虚拟场景生成等计算机视觉任务。