seetaface6ToPy

1.简介

项目基于SeetaFace6 官方地址为:https://github.com/SeetaFace6Open/index

灵感来源:https://github.com/tensorflower/seetaFace6Python

由于上面的版本没有提供FaceAPI的源码,而且不支持GPU,且功能并没有写全, 故自己利用ctypes编写了一个API.由于我的电脑windows系统编译seetaface老是出错,原因是电脑安装了VS又有MinGW,编译会自动寻找vs,但是又提示在编译某些文件时'cl.exe'找不到,可是我已经设置过环境变量了,百度了很多方案,都不行。(另外这个错误在安装insightface最新版本也出现,如果有知道解决方案的请告诉我)。所以放弃了windows版本,只编译了linux的.linux版本不区分centos跟ubuntu,都可以使用.设置方法一样.

2.模型下载

模型可以在seetaface的官网下载:

模型文件： Part I: Download code: ngne, including: age_predictor.csta, face_landmarker_pts5.csta, fas_first.csta, pose_estimation.csta, eye_state.csta, face_landmarker_pts68.csta, fas_second.csta, quality_lbn.csta, face_detector.csta, face_recognizer.csta, gender_predictor.csta, face_landmarker_mask_pts5.csta, face_recognizer_mask.csta, mask_detector.csta. Part II: Download code: t6j0，including: face_recognizer_light.csta.

3.编译

cd seetaface
mkdir build && cd build
cmake ..
make

将在lib文件夹下产生libSeetaFaceAPI.so文件

4. 设置动态库到环境中

sudo echo  ${seetaFace6ToPy目录路径}/seetaface/lib/ > /etc/ld.so.conf.d/seetaface6.con  
sudo ldconfig

5.运行seeta_test.py

使用说明:

1.把需要调用的功能放进一个列表中,名称必须参照下面:

func_list =["FACE_DETECT","LANDMARKER5","LIVENESS","LANDMARKER_MASK","FACE_AGE","FACE_GENDER","FACE_RECOGNITION","MOUTH_MASK","EYE_STATE","FACE_CLARITY","FACE_BRIGHT","FACE_RESOLUTION","FACE_POSE","FACE_INTEGRITY", "FACE_TRACK"]

2.设置使用cpu/gpu,并初始化

model_path = "./seeta/model"
seetaFace = SeetaFace(func_list,device=0,id=0)
seetaFace.init_engine(model_path)

device: ,自动:0, cpu:1, gpu:2 (经过实践:自动并不自动,只会调cpu)

id:处理器的编号,如果为cpu,为0就可以.

6.如果觉得有用,点个小星星吧.