fregulationn / python-REST

standrard python REST service

Face_fusion_rest

Installation

1. Get the code. We will call the cloned directory as $FACE_FUSION_ROOT.

git clone https://github.com/fregulationn/python-REST.git

2. Build the code. Please follow FaceNet to install all necessary packages and build it.We will call FaceNet cloned directory as $FACENET_ROOT.

cd $FACE_FUSION_ROOT
pip install -r requirements.txt
export PATH=$PATH:FACENET_ROOT

Pre-trained models

Model name	LFW accuracy	Training dataset	Architecture
20180408-102900	0.9905	CASIA-WebFace	Inception ResNet v1
20180402-114759	0.9965	VGGFace2	Inception ResNet v1

Interface

1.根据OpenID检查用户是否存在, 不存在则添加新用户

URL: checkUser
Type: POST
body：   
{
  openId:'qk125463'
}

:return 
{
  status:True/False  
}

2.人脸融合

URL: fusion
Type: POST
body：   
{
    user_Id:'qk125463'
    inputImage:
}

返回值：
{
    user_Id:'qk125463',
    type:'fusion',
    time:'Fri, 19 Apr 2019 20:12:21 GMT',
    outputImage:
}

3.人脸检测

URL: detect
Type: POST
body：   
{
  user_Id:'qk125463'
  inputImage: 
}
返回值：
{
      user_Id:'qk125463',
      type:'detect',
      time:'Fri, 19 Apr 2019 20:12:21 GMT',
      outputImage:
}

4.人脸匹配

URL: recognize
Type: POST
body：   
{
  user_Id:'qk125463'
  inputImage:
}

返回值：
{
  user_Id:'qk125463',
  type:'recognize',
  time:'Fri, 19 Apr 2019 20:12:21 GMT',
  outputImage:
}

5.保存(TODO)

URL: save
Type: POST
body：   
{
  user_Id:'qk125463',
  history_id:'1',
  type:'fusion/detect/recognize',
  time:'Fri, 19 Apr 2019 20:12:21 GMT',
  outputImage:
}
返回值：
  true/false

6.获取历史数据

URL: user/getHistory
Type: POST
body：   
{
  openId:'qk125463'
}
返回值：

{ 
  history：
  [
    {
      user_Id:'qk125463',
      history_id:'1',
      type:'fusion',
      time:'Fri, 19 Apr 2019 20:12:21 GMT',
      outputImage:
    },
    {
      user_Id:'qk125463',
      history_id:'1'
      type:'detect',
      time:'Fri, 19 Apr 2019 20:12:21 GMT',
      outputImage:
    }
  ] 
}

Dataset

1.User 
  id 
  username 

2.Log  
  id 
  username 
  datatime  
  imageres 
  type: 'fusion','detect','recognize' 

3.Image 
  id  
  imagepath  
  feature  
  （把这个浮点数向量使用python的json模块进行序列化 json.dumps 成为一个字符串后以TEXT类型数据存储，取出的时候再使用json.load还原成向量，浮点数精度取了10位，粗略估计一下按20计算每一个维度，则每一个向量存储空间不大于20*128，TEXT类型能够存储下 [出处](https://www.jianshu.com/p/eead9790ea97)）

Detect && Recognize

检测和识别使用的框架来自FaceNet，出处，检测所使用的方法为MTCNN，识别为FaceNet，详情见出处