Aesthetic mixed dataset with attributes(AMD-A)
收藏资源简介:
AMD-A数据集是由北京电子科技学院构建的一个包含16924张图像的大型美学属性评估数据集。该数据集通过混合多个现有数据集并设计外部属性特征,旨在解决传统美学评估数据集标签有限和数据量少的问题。AMD-A数据集分为两部分,一部分用于美学总体评分回归,另一部分用于美学属性评分分类和回归。数据集中的每张图像都标注了光、色、构图三个属性标签和一个总体标签。该数据集的应用领域主要集中在图像美学质量评估,通过提供更丰富的属性标签和更大的数据量,支持开发更精确的美学评估模型,以模拟人类对图像美学的感知和评价。
The AMD-A dataset is a large-scale aesthetic attribute evaluation dataset constructed by Beijing Electronic Science and Technology Institute, comprising 16,924 images. By integrating multiple existing datasets and designing external attribute features, it is developed to address the limitations of limited labels and small data scale in traditional aesthetic evaluation datasets. The AMD-A dataset is split into two subsets: one for overall aesthetic score regression, and the other for aesthetic attribute score classification and regression. Every image in the dataset is annotated with three attribute labels (lighting, color, and composition) and one overall aesthetic label. Its primary application scenarios center on image aesthetic quality assessment. By offering richer attribute labels and a larger-scale dataset, it enables the development of more accurate aesthetic evaluation models that simulate human perceptual judgment and evaluation of image aesthetics.




