AI生成图片数据集
收藏资源简介:
本数据集面向多模态大模型训练团队、AIGC图像质量评估研究机构及计算机视觉算法开发者,旨在解决AI合成图像的元数据结构化程度不足、生成参数与图像内容难以关联分析的问题。数据集包含即梦AI平台生成的图片及对应的元数据,元数据以JSON格式存储,记录生成参数、模型版本、提示词、生成时间等关键信息。与传统仅提供生成图像的做法不同,本数据集将“创作过程信息”与图像共同封装,使研究者可追溯每张图像的生成来源与技术参数。
This dataset is designed for multimodal large language model (LLM) training teams, AIGC image quality assessment research institutions, and computer vision algorithm developers. It aims to solve the issues of insufficient structuralization of metadata for AI-synthesized images and the challenge of correlating generation parameters with their corresponding image content for joint analysis. The dataset comprises images generated on the Jimeng AI Platform and their corresponding metadata, which is stored in JSON format and records critical information including generation parameters, model version, prompts, and generation time. Unlike conventional datasets that only provide synthesized images, this dataset encapsulates the "creation process information" alongside the images, allowing researchers to trace the generation source and technical parameters of each image.




