Face alignment using a deep neural network with local feature learning and recurrent regression
作者：
Highlights：
• We generate a global feature map using a network that trained a local feature on face.
• Using a local feature extraction layer, local features are selectively investigated.
• Landmark positions are estimated via regression on the extracted feature recurrently.
• Extracted features from generated global feature map show distinctive property.
• Face alignment via proposed method shows a state-of-the-art result on public dataset.
摘要
•We generate a global feature map using a network that trained a local feature on face.•Using a local feature extraction layer, local features are selectively investigated.•Landmark positions are estimated via regression on the extracted feature recurrently.•Extracted features from generated global feature map show distinctive property.•Face alignment via proposed method shows a state-of-the-art result on public dataset.
论文关键词：Face alignment,Deep neural network,Convolutional neural network,Local feature learning,Head pose estimation,Facial landmark tracking
论文评审过程：Received 11 February 2017, Revised 11 June 2017, Accepted 12 July 2017, Available online 13 July 2017, Version of Record 26 July 2017.
论文官网地址：https://doi.org/10.1016/j.eswa.2017.07.018
