Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (9): 2741-2747.DOI: 10.11772/j.issn.1001-9081.2020111847

Special Issue: 多媒体计算与计算机仿真

• Multimedia computing and computer simulation • Previous Articles     Next Articles

Facial landmark detection based on ResNeXt with asymmetric convolution and squeeze excitation

WANG Hebing, ZHANG Chunmei   

  1. School of Computer Science and Engineering, North Minzu University, Yinchuan Ningxia 750021, China
  • Received:2020-11-25 Revised:2021-02-26 Online:2021-09-10 Published:2021-09-15
  • Supported by:
    This work is partially supported by the General Project of Ningxia Hui Autonomous Region Key Research and Development Plan (2019BDE0311), the First-Class Discipline Construction (Electronic Science and Technology Discipline) Funded Project in Ningxia Colleges and Universities (NXYLXK2017A07).


王贺兵, 张春梅   

  1. 北方民族大学 计算机科学与工程学院, 银川 750021
  • 通讯作者: 张春梅
  • 作者简介:王贺兵(1992-),男,河北邢台人,硕士研究生,主要研究方向:深度学习、人脸关键点检测;张春梅(1964-),女,宁夏银川人,教授,硕士,CCF会员,主要研究方向:视觉信号处理、遥感图像处理。
  • 基金资助:

Abstract: Cascaded Deep Convolutional Neural Network (DCNN) algorithm is the first model that uses Convolutional Neural Network (CNN) in facial landmark detection and the use of CNN improves the accuracy significantly. This strategy needs to perform regression processing to the data between the adjacent stages repeatedly, resulting in complex algorithm procedure. Therefore, a facial landmark detection algorithm based on Asymmetric Convolution-Squeeze Excitation-Next Residual Network (AC-SE-ResNeXt) was proposed with only single-stage regression to simplify the procedure and solve the non-real-time problem of data preprocessing between adjacent stages. In order to keep the accuracy, the Asymmetric Convolution (AC) module and the Squeeze-and-Excitation (SE) module were added to Next Residual Network (ResNeXt) block to construct the AC-SE-ResNeXt network model. At the same time, in order to fit faces in complex environments such as different illuminations, postures and expressions better, the AC-SE-ResNeXt network model was deepened to 101 layers. The trained model was tested on datasets BioID and LFPW respectively. The overall mean error rate of the model for the five-point facial landmark detection on BioID dataset was 1.99%, and the overall mean error rate of the model for the five-point facial landmark detection on LFPW dataset was 2.3%. Experimental results show that with the simplified algorithm procedure and end to end processing, the improved algorithm can keep the accuracy as cascaded DCNN algorithm, while has the robustness significantly increased.

Key words: facial landmark detection, Asymmetric Convolution (AC), Squeeze-and-Excitation (SE) module, Convolutional Neural Network (CNN), Next Residual Network (ResNeXt)

摘要: 级联深度卷积神经网络(DCNN)算法为首先在人脸关键点检测中使用卷积神经网络(CNN)的模型,CNN的使用使得检测精度得到极大的提升。针对该策略需要对相邻阶段间的数据反复进行回归处理使得算法流程十分复杂的问题,提出基于非对称卷积-压缩激发-次代残差网络(AC-SE-ResNeXt)的人脸关键点检测算法。所提算法仅使用单阶段回归,既避免了级联策略中多阶段回归的算法流程复杂性,又解决了相邻阶段间数据需要进行预处理的问题。为了不降低精度,在次代残差网络(ResNeXt)块的基础上添加了非对称卷积(AC)模块和压缩激发(SE)模块,构建了AC-SE-ResNeXt网络模型。同时,为了能够精确拟合在不同光照、姿态、表情等复杂环境下的人脸,将AC-SE-ResNeXt网络模型加深到101层。对训练好的模型分别在数据集BioID和LFPW上进行测试,其中该模型在BioID数据集上的人脸五点关键点检测的综合平均误差率为1.99%,在LFPW数据集上的人脸五点关键点检测的综合平均误差率为2.3%。实验结果表明,所改进的算法不但简化了算法流程使之能进行端到端处理,而且其精度与级联DCNN算法相当,鲁棒性也有明显提升。

关键词: 人脸关键点检测, 非对称卷积, 压缩激发模块, 卷积神经网络, 次代残差网络(ResNeXt)

CLC Number: