Style Transfer Dataset
收藏资源简介:
Style Transfer Dataset 是由莫斯科国立大学的研究人员创建的一个用于图像风格迁移研究的数据集。该数据集包含10,000个手动评级的风格化图像,涵盖了多种内容和风格图像,旨在帮助研究人员评估和优化风格迁移算法。数据集的创建过程包括选择多样化的内容和风格图像,使用ArtFlow算法生成风格化图像,并通过三名标注者对每个风格化图像进行1-10分的评分。该数据集可用于自动化风格迁移配置和评估,特别是在设计、时尚和广告等领域中应用广泛。
The Style Transfer Dataset is a research dataset for image style transfer studies, created by researchers from Lomonosov Moscow State University. It contains 10,000 manually rated stylized images derived from a diverse range of content and style source images, and is designed to assist researchers in evaluating and optimizing style transfer algorithms. The dataset's development process includes selecting varied content and style reference images, generating stylized images using the ArtFlow algorithm, and having three annotators score each stylized image on a 1-to-10 scale. This dataset can be applied to automated style transfer configuration and evaluation, and is widely used particularly in fields such as design, fashion, and advertising.




