Open joyhuang9473 opened 7 years ago
(b) We create a guided by 2D landmarks network which converts 2D landmark annotations to 3D and unifies all existing datasets, leading to the creation of LS3D-W, the largest and most challenging 3D facial landmark dataset to date (~230,000 images)
【世界最大人脸对齐数据集】ICCV 2017:距离解决人脸对齐已不远 https://mp.weixin.qq.com/s/s5HL6y2P9_KqpSAQg08URw
How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)