使用face_recognition进行人脸特征检测

淩亂°似流年 2021-12-24 06:13 305阅读 0赞

效果图  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70]  
调用face\_recognition.face\_landmarks()方法即可得到人脸特征点, 返回一个字典, 下图是返回的数据, 包括chin(下巴), left\_eye(左眼)等.  
![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70 1]  
我画了两种图, 一种是遍历所有的点, 直接给点画图的图(点用实心圆绘制). 第二个是单独画下巴, 连成线, 用的是polylines方法.

我是4.10版本的opencv. 查阅官方py文档, [这是链接][Link 1]![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70 2]  
完整代码:

import face_recognition
    import numpy as np
    import cv2
    
    image = face_recognition.load_image_file("./data/奥巴马.png")
    image2 = image.copy()
    face_landmarks_list = face_recognition.face_landmarks(image)
    # print(face_landmarks_list)
    for each in face_landmarks_list:
        print(each)
        for i in each.keys():
            print(i, end=': ')
            print(each[i])
            for any in each[i]:
                image = cv2.circle(image, any, 3, (0,0,255), -1)
    cv2.imshow("奥巴马", image)
    
    # 单独画下巴
    for each in face_landmarks_list:
        pts = np.array(each['chin'])
        pts = pts.reshape((-1, 1, 2))
        cv2.polylines(image2, [pts], False, (0, 255, 255)) # false 参数使其不闭合
    cv2.imshow("奥巴马2", image2)
    
    cv2.waitKey(0)
    cv2.destroyAllWindows()

在线摄像机版本:

import face_recognition
    import numpy as np
    import cv2
    
    camera = cv2.VideoCapture(0)
    
    while True:
        ret, image = camera.read()
        image = cv2.flip(image, 1)
        image2 = image.copy()
    
        face_landmarks_list = face_recognition.face_landmarks(image)
        # print(face_landmarks_list)
        for each in face_landmarks_list:
            print(each)
            for i in each.keys():
                print(i, end=': ')
                print(each[i])
                for any in each[i]:
                    image = cv2.circle(image, any, 3, (0,0,255), -1)
        cv2.imshow("奥巴马", image)
    
        # 单独画下巴
        for each in face_landmarks_list:
            pts = np.array(each['chin'])
            pts = pts.reshape((-1, 1, 2))
            cv2.polylines(image2, [pts], False, (0, 255, 255)) # false 参数使其不闭合
        cv2.imshow("奥巴马2", image2)
        if cv2.waitKey(1000 // 12) & 0xff == ord("q"):
            break
    
    cv2.destroyAllWindows()
    camera.release()

附一份在线的人脸搜索代码, 人脸数据保存在相对路径`./data/mans` 下

import cv2
    import face_recognition
    import numpy as np
    import os
    import re
    
    # 人脸数据, 文件, 编码, 名字
    files = os.listdir("./data/mans")
    face_images = [0]*len(files)
    face_encodings = [0]*len(files)
    face_names = [0]*len(files)
    
    # 获取编码和名称
    for i in range(len(files)):
        face_images[i] = face_recognition.load_image_file('./data/mans/' + files[i])
        face_encodings[i] = face_recognition.face_encodings(face_images[i])
        if len(face_encodings[i]) > 0:
            face_encodings[i] = face_encodings[i][0]
        else:
            face_encodings[i] = None
        face_names[i] = re.findall(r'(.*)\..*', files[i])[0]
    print(face_names)
    
    # 人脸比较
    # results = face_recognition.compare_faces(face_encodings[0], face_encodings[1])
    # print(results)
    
    # 人脸距离
    # face_distances = face_recognition.face_distance(face_encodings[0], face_encodings[1])
    # index = np.argmin(face_distances)
    # print(index)
    
    # camera = cv2.VideoCapture('./data/test.avi') # 从视频文件
    camera = cv2.VideoCapture(0) # 从摄像头
    
    while True:
        ret, img = camera.read()
        img = cv2.flip(img, 1)
        # img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 灰度处理
        locations = face_recognition.face_locations(img)
        for top, right, bottom, left in locations:
            cv2.rectangle(img, (left, top), (right, bottom), (255, 0, 0), 2)
            sub_img = img[top:bottom, left:right]
            sub_img_code = face_recognition.face_encodings(sub_img)
            if len(sub_img_code) != 0:
                face_distances = face_recognition.face_distance(face_encodings, sub_img_code[0])
                print(face_distances)
                index = np.argmin(face_distances)
                name = face_names[index]
                cv2.putText(img, name, (left, top - 20), cv2.FONT_HERSHEY_SIMPLEX, 1, 255, 2)
        cv2.imshow('Face', img)
        if cv2.waitKey(1000 // 12) & 0xff == ord("q"):
            break
    
    cv2.destroyAllWindows()
    camera.release()

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70]: https://img-blog.csdnimg.cn/20190626202632765.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc=,size_16,color_FFFFFF,t_70
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70 1]: /images/20211223/5224e3d4ed7b4cbf9928e60c8bbe88ac.png
[Link 1]: https://docs.opencv.org/4.1.0/dc/da5/tutorial_py_drawing_functions.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dqaDI2MjIwNzUxMjc_size_16_color_FFFFFF_t_70 2]: /images/20211223/6ebf486e38b9417ca2b25353fb664126.png