WFLW 人脸数据集

WFLW 全称 Wider Facial Landmarks in the Wild，包含 10,000 张脸（7,500 张用于训练，2,500 张用于测试），共 98 个标注的 landmark 。除 landmark 外，该数据集还包含丰富的属性标注（遮挡，姿态，妆容，光照， 模糊和表情等）。

WFLW, short for Wider Facial Landmarks in the Wild, consists of 10,000 facial images (7,500 for training and 2,500 for testing) with a total of 98 annotated facial landmarks. In addition to the landmarks, this dataset also includes rich attribute annotations such as occlusion, pose, makeup, illumination, blurriness, expression and more.

搜集汇总

背景与挑战

背景概述

WFLW人脸数据集全称为Wider Facial Landmarks in the Wild，包含10,000张人脸图像，其中7,500张用于训练、2,500张用于测试。每张图像标注了98个关键点，并提供了遮挡、姿态等丰富的属性信息。