Main > Laser & Optoelectronics Progress >  Volume 57 >  Issue 21 >  Page 210703 > Article
  • Abstract
  • Abstract
  • View Summary
  • Figures (5)
  • Tables (4)
  • Equations (0)
  • References (20)
  • Get PDF(in Chinese)
  • Paper Information
  • Received: Jan. 8, 2020

    Accepted: --

    Posted: Nov. 1, 2020

    Published Online: Oct. 27, 2020

    The Author Email: Jiajia Jiang (

    DOI: 10.3788/LOP57.210703

  • Get Citation
  • Copy Citation Text

    Xu Lihuai, Li Zhe, Jiang Jiajia, Duan Fajie, Fu Xiao. High-Precision and Lightweight Facial Landmark Detection Algorithm[J]. Laser & Optoelectronics Progress, 2020, 57(21): 210703

    Download Citation

  • Category
  • Fourier Optics and Signal Processing
  • Share
Laser & Optoelectronics Progress, Vol. 57, Issue 21, 210703 (2020)

High-Precision and Lightweight Facial Landmark Detection Algorithm

Lihuai Xu1, Zhe Li2, Jiajia Jiang1,*, Fajie Duan1, and Xiao Fu1

Author Affiliations

  • 1天津大学精密测试技术及仪器国家重点实验室, 天津 300072
  • 2中国科学院深海科学与工程研究所, 海南 三亚 572000


In view of the high complexity of the current facial landmark detection algorithm network model, which is not conducive to deployment on devices with limited computing resources, this paper proposes a high-precision and lightweight facial landmark detection algorithm based on the idea of knowledge distillation. This algorithm improves the Bottleneck module of residual network(ResNet50) and introduces packet deconvolution to obtain a lightweight student network. At the same time, a pixel-wise loss function and a pair-wise loss function are proposed. By aligning the output feature maps and intermediate feature maps of the teacher network and the student network, the prior knowledge of the teacher network is transferred to the student network, thereby improving the detection accuracy of the student network. Experiments show that the student network obtained by this algorithm has only 2.81M parameter amount and 10.20MB model size, the frames per second on the GTX1080 graphics card is 162frames and the normalized mean error on 300W and WFLW datasets are 3.60% and 5.50%, respectively.


Please Enter Your Email: