dlib_face_API

dlib官网

环境配置

windows 系统

下载安装cmake
pip安装 pip install dlib

ubuntu系统

命令安装cmake sudo apt install cmake
pip安装 pip install dlib

快速启动

本工程ubuntu下能够自动下载模型参数并解压。
由于bz2在windows上不便脚本解压，因此windows系统请先去dlib下载目录下载模型文件并放置在modules/face_server/params文件夹中。
运行之前需要调用新建数据库的api，详见modules/utils/api.py中的new_database()方法
注意人脸注册图片中有且仅有一张脸，否则视为无效注册，数据库不做更新。

需要开启http和websocket两个服务

运行app.py
websocket_serve.py中start_server = websockets.serve(time, "10.20.50.163", 5678)ip地址更换为客户端地址
运行websocket_serve.py

本服务端已具备人脸注册、信息删除或修改、已注册人脸信息查询等功能。前端调用对应接口即可

flask 服务负责数据库增删改查功能

人脸信息注册

@app.route('/uploadInfo', methods=["POST"])
输入输出详见 api.py 文档中的 face_register 方法

人脸更新（有图片更新）

@app.route('/updateInfo', methods=["POST"])
输入输出详见 api.py 文档中的 face_update 方法

人脸更新（无图片）

@app.route('/updateInfo2', methods=["POST"])
输入输出详见 api.py 文档中的 face_update 方法

人脸删除

@app.route('/deleteInfo', methods=["POST"])
输入输出详见 api.py 文档中的 face_delete 方法

人脸信息获取

@app.route('/getAllInfo', methods=["POST"])
输入输出详见 api.py 文档中的 face_get_info 方法

websocket负责图像帧实时传输

start_server = websockets.serve(time, "10.20.50.163", 5678)

输入输出详见 api.py 文档中的 search_identity 方法

api.py 文档

Help on module modules.utils.api in modules.utils:

NAME

modules.utils.api

DESCRIPTION

File Name：     faceRecognition.py
Description :  Instantiate FaceRecognition from faceRecognition.py. Give a clear API for user.
Author :       KangJiaHui
date：         2020/11/26

FUNCTIONS

face_delete(user_id)
    Delete an exact face_info by user_id
    :param user_id: string e.x. "10098440"
    :return: None, the face_info will be deleted from database.

face_get_info(page_num=1, max_rows=100)
    Fetch face_info page by page. You can only fetch one page at once.
    :param page_num: int, page index
    :param max_rows: int, how many registered faces in one page
    :return: dict, e.x. {"result": 0, "message": "SUCCESS",
                        "user_data": [{"userID": "10098440", "userGroup": "staff", "userGender": "女", "userName": "康佳慧",
                                        "latest_modify_time": "2020-12-15 15:15:41", "userIMG": base64_image}, ]}

face_register(input_dict)
    Registers only one picture.
    :param input_dict:
        e.x. {"user_id": "10098440", "group_id": "staff", "gender": "女", "user_info": "康佳慧", "user_image": base64_image}
    :return: None, results will be written into database.

face_update(input_dict)
    Update exist face_info.
    :param input_dict: e.x. {"user_id": "10098440", "group_id": "staff", "user_info": "康佳慧", "user_image": base64_image}
    :return: None, results will be written into data_path.

new_database()
    Initial a new database. Only execute once before use.
    :return: None. Database and table will be established.

search_identity(image=None, path=None, thresh=0.4)
    Serch all faces in one image in the register of n faces' features.
    If path and image coexist, then path will cover image.
    :param image: numpy.ndarray
    :param path: str, indicates an image
    :param thresh: distance between face and matched face should be smaller than thresh
    :return:dict, in format of
            {"result": 0, "message": "SUCCESS", "image": encoded_base64_img,
            "faces":[{"user_id": "10098440", "group_id": "staff", "gender": "女", "user_info": "康佳慧",
                    "box": [216, 118, 439, 341], "distance": 0.35670, "image": base64_image}, ]}

DATA

face = <modules.face_server.faceRecognition.FaceRecognition object>

FILE

e:\deeplearning\kang\face_api\modules\utils\api.py