EDFace-Celeb-1M

EDFace-Celeb-1M是一个大规模的公开人脸幻觉数据集，由澳大利亚国立大学工程与计算机科学学院的研究人员创建。该数据集包含170万张来自不同国家和种族的名人照片，旨在解决现有数据集在种族分布上的不平衡问题。数据集内容丰富，包括不同分辨率的人脸图像，支持多种放大倍数（如2×, 4×, 8×），并提供固定训练和测试集，以便于公平比较和重复性研究。此外，数据集还设计了多种评估协议，用于评估和推动人脸幻觉算法的发展，适用于人脸分析等多个领域。

EDFace-Celeb-1M is a large-scale open-access face hallucination dataset developed by researchers from the School of Engineering and Computer Science at the Australian National University. This dataset comprises 1.7 million celebrity portraits sourced from various countries and ethnic groups, with the objective of mitigating the racial distribution imbalance present in existing face datasets. The dataset offers rich content, including face images of diverse resolutions, supports multiple magnification factors (e.g., 2×, 4×, 8×), and provides fixed training and test splits to enable fair comparative evaluation and reproducible research. Additionally, it incorporates multiple standardized evaluation protocols designed to assess and advance the development of face hallucination algorithms, rendering it applicable to multiple domains including face analysis.