中国移动通用场景图文数据集

本数据集为通用场景图文配对数据集，涵盖日常生活、自然景观、艺术时尚等各类场景的图像及对应文本描述、核心元素标签，覆盖人物、美食、风景、建筑、城市、艺术等多类别内容。样本描述语言支持中文与英文，同步提供图像主要元素标注标签，具备强场景属性、图文对应属性与多语言属性，适合用于多模态大模型训练、图文一致性校验、图像分类、文生图模型优化、智能标签生成、图像内容检索及视觉内容审核。This is a general-scenario image-text pairing dataset. It covers images from various scenarios such as daily life, natural landscapes, art and fashion, along with their corresponding text descriptions and core element tags, and encompasses diverse categories including people, food, scenery, architecture, cities, art and more. The sample descriptions support both Chinese and English, and annotated tags for the main elements of the images are also provided. This dataset features strong scenario relevance, accurate image-text alignment and multilingual capabilities, making it suitable for training multimodal large language models, image-text consistency verification, image classification, text-to-image model optimization, intelligent tag generation, image content retrieval and visual content auditing.

搜集汇总

背景与挑战

背景概述

该数据集是一个通用场景图文配对数据集，包含日常生活、自然景观、艺术时尚等多类场景的图像及其对应的中英文文本描述与核心元素标签。它具备强场景、图文对应和多语言属性，适用于多模态大模型训练、图文一致性校验、图像分类等多种人工智能应用。