FFHQ-Ref
收藏资源简介:
FFHQ-Ref数据集由联发科技和国立阳明交通大学创建,包含20,405张高质量人脸图像及其对应的参考图像。该数据集旨在为基于参考图像的人脸修复模型提供训练和评估数据。数据集的创建过程包括使用ArcFace模型进行人脸识别,以确定同一个人的图像,并将其分组。FFHQ-Ref数据集主要应用于人脸图像修复领域,旨在解决低质量图像中人脸特征不准确的问题。
The FFHQ-Ref dataset was created by MediaTek and National Yang Ming Chiao Tung University, which contains 20,405 high-quality facial images and their corresponding reference images. This dataset is designed to provide training and evaluation data for reference-based face restoration models. The dataset creation process applies the ArcFace model for face recognition to identify images of the same individual and group them accordingly. The FFHQ-Ref dataset is primarily utilized in the field of facial image restoration, aiming to address the issue of inaccurate facial features in low-quality images.




