A Creative Industry Image Generation Dataset Based on Captions and Sketches
收藏arXiv2022-11-17 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2211.09035v1
下载链接
链接失效反馈官方服务：
资源简介：
本数据集是首个覆盖创意产业四大重要领域的数据集，包括概念设计、平面设计、3D-CG和户外设计，并标注有提示和草图。数据集从CC12M等文本视觉数据集中选取属于创意产业领域的图像，使用CLIPasso模型生成草图，并通过人工筛选确保草图质量。该数据集旨在解决图像生成方法在创意产业中的可控性问题，提供多参考图像和细粒度评分，以更准确地评估图像生成模型的效果。
This dataset is the first-of-its-kind resource covering four core domains of the creative industry: conceptual design, graphic design, 3D-CG, and outdoor design, and is annotated with prompts and sketches. Images falling under the creative industry domain were curated from text-visual datasets such as CC12M during dataset construction. Sketches were generated using the CLIPasso model, with their quality validated via manual screening. This dataset aims to resolve the controllability challenge of image generation methods within the creative industry, and provides multi-reference images and fine-grained scoring metrics to enable more accurate evaluation of image generation model performance.
提供机构：
字节跳动
创建时间：
2022-11-17
