FFHQ-UV
收藏资源简介:
FFHQ-UV数据集是由南京理工大学和腾讯AI实验室合作创建的大规模人脸UV纹理数据集,包含超过50,000个高质量的UV纹理图,这些纹理图具有均匀的照明、中性表情和清洁的面部区域,非常适合用于不同光照条件下的真实感渲染。该数据集通过自动化的流程从FFHQ数据集中提取,利用StyleGAN技术生成多视角标准化的人脸图像,并通过精细的UV纹理提取、校正和完成过程,确保纹理图的质量。FFHQ-UV数据集的应用领域包括3D人脸重建,旨在提高重建的准确性和纹理图的质量,使其更适合实际应用。
FFHQ-UV dataset is a large-scale facial UV texture dataset jointly developed by Nanjing University of Science and Technology and Tencent AI Lab. It contains over 50,000 high-quality UV texture maps featuring uniform illumination, neutral expressions and clean facial regions, which are highly suitable for photorealistic rendering under diverse lighting conditions. This dataset is extracted from the original FFHQ dataset through an automated pipeline: it first leverages the StyleGAN technique to generate multi-view standardized facial images, then performs elaborate UV texture extraction, correction and completion processes to guarantee the quality of the texture maps. Applications of the FFHQ-UV dataset cover 3D face reconstruction, where it aims to improve the reconstruction accuracy and the quality of texture maps, making the technology more suitable for real-world applications.




