ArtBench-10
收藏资源简介:
ArtBench-10是由卡内基梅隆大学和加州大学伯克利分校的研究人员共同创建的一个艺术风格多样、类别平衡的高质量数据集。该数据集包含60,000张艺术品图像,涵盖10种不同的艺术风格,每种风格包含5,000张训练图像和1,000张测试图像。数据集通过严格的数据收集、标注、过滤和预处理流程创建,确保图像质量和标签的准确性。ArtBench-10适用于评估和比较不同艺术风格生成模型的性能,为艺术生成领域的研究提供了一个标准化的基准。
ArtBench-10 is a high-quality, class-balanced dataset featuring diverse art styles, jointly created by researchers from Carnegie Mellon University and the University of California, Berkeley. It contains 60,000 artwork images covering 10 distinct art styles, with 5,000 training images and 1,000 test images for each style. The dataset is constructed through rigorous data collection, annotation, filtering and preprocessing pipelines, ensuring the quality of images and the accuracy of its labels. ArtBench-10 is suitable for evaluating and comparing the performance of different art style generation models, providing a standardized benchmark for research in the field of art generation.

- 1The ArtBench Dataset: Benchmarking Generative Models with Artworks卡内基梅隆大学 · 2022年



