ImagenHub Benchmark Dataset

该数据集包含了来自不同类别（包括动物和日常物品）的独特主体的样本。这些图像取自DreamBooth，在所有实验中的分辨率均为512x512。具体来说，该数据集包含了154个样本，涉及22个独特的主体用于编辑任务；同时，还包含了150个提示，涉及29个独特主体用于生成任务。这些任务的目的是实现主体驱动的图像编辑和图像生成。

This dataset contains samples of unique subjects from various categories, including animals and everyday objects. All images in the dataset are sourced from DreamBooth, with a fixed resolution of 512×512 across all experiments. Specifically, the dataset includes 154 samples involving 22 unique subjects for image editing tasks; additionally, it contains 150 prompts related to 29 distinct subjects for image generation tasks. The goals of these tasks are subject-driven image editing and image generation.