Skip to content

face alignment of state-of-art implemented with tnsorflow

Notifications You must be signed in to change notification settings

Susia00/faceAlignment_tensorflow

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Facial Alignment

faceAlignment_tensorflow 基于tensorflow 实现了现有先进的人脸对齐算法(the state-of-the-art deep learning models) 主要包括以下几个方面:

  1. 利用tensorfow复现了现有人脸对齐算法的训练代码,尽量达到paper中的水平
  2. 基于numpy实现了人脸对齐领域常用的几种评价指标(如MSE, normalized MSE, etc.)
  3. 基于tf.data API 实现了对不同标注格式的人脸对齐数据集的加载,预处理和可视化
  4. 简单的数据增广

依赖

  • tensorflow >=1.4.0
  • python == 3.5
  • opencv == 3.3.0
  • numpy

组织结构

  • face_alignment: 根目录
    • model_zoo: 定义算法网络模型文件
    • tools: 训练,评测,数据可视化,数据增广脚本
    • utils:辅助功能,如绘制点/矩形框,数据集加载,日志,评测指标等
  • model: 训练模型存放路经
  • data 相关数据资源文件,如meanshape等
  • README.md

支持的算法

支持的数据集

  • 300W (image-pts pair)
  • 300W-LP(image-mat pair)
  • AFLW2000-3D(image-mat pair)
  • 300VW(todo)

性能对比

  • MSE normalized by pupil distance
网络 stage 300W-common 300W-challege AFLWW2000 speed(ms/face) 显存占用(%)
dan_vgg_112_300W 1/2 1 14 14 1.3 65
dan_vgg_112_300WAugment 1/2 7.43/6.10 14.99/11.40 49.013/40.23 3-5 25
dan_resnet_112_300WAugment 1/2 8.933/0.06133 13.430/11.345 44.712/44.072 4-7
dan_mobilenet_112_300WAugment 1/2 8.323/8.294 14.892/14.617 48.134/47.506 2-5
prnet_256_300WLP - 7.607 14.631 14.988 10 75
dan_resnet_112_300WAugment 1/2 1 32 8 1.3 78
  • MSE normalized by diagonal box distance
网络 stage 300W-common 300W-challege AFLWW2000 speed(ms/face) 显存占用(%)
prnet_256_300WLP(paper) - 2.22 3.67 2.3 10 -
FAN_256_300WLP(paper) - - - 3.38 10 -
FAN_256_300WLP(train) - 2.15 3.68 2.57 25 -

About

face alignment of state-of-art implemented with tnsorflow

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%