赞
踩
首先,我是考虑,这系统在Windows下做还是在Linux、Ubuntu下做比较好?
然后,我都检测过,Windows下可以用python、anaconda写代码都可以。当然,和在Linux和Ubuntu下使用的库和包肯定也是一样的。目的就在于那种形式,比较简单,这是我的想法。何必给自己添加难度呢?
后来,我是选择在Ubuntu下做人脸识别。至于为什么,简单是一回事,后面还有原因,我会一一讲解的。
我写了一份草图,我这几周分别做哪些事。
搭建环境
https://pypi.tuna.tsinghua.edu.cn/simple
sudo apt-get update
sudo apt-get upgrade
需要的包的大概有:numpy、opencv-python、keras、scikit-learn、tensorflow、dlib、pandas(有条件可以考虑基于GPU的)以及神经网络所需的各种包及库等等
sudo pip install opencv-python
sudo pip install keras
sudo pip install scikit-learn
sudo pip install tensorflow
sudo pip install dlib
sudo pip install pandas
主要分为五个模块:
1.人脸收集
2.头像提取
3.尺寸变换
4.训练
5.识别
话不多说上代码:
测试摄像头
文件名为:how_to_use_camera.py
这里说明下,小编都是用python来编译的
# OpenCv 调用摄像头 # 默认调用笔记本摄像头 # Author: coneypo # Blog: http://www.cnblogs.com/AdaminXie # GitHub: https://github.com/coneypo/Dlib_face_recognition_from_camera # Mail: coneypo@foxmail.com import cv2 cap = cv2.VideoCapture(0) # cap.set(propId, value) # 设置视频参数: propId - 设置的视频参数, value - 设置的参数值 cap.set(3, 480) # cap.isOpened() 返回 true/false, 检查摄像头初始化是否成功 print(cap.isOpened()) # cap.read() """ 返回两个值 先返回一个布尔值, 如果视频读取正确, 则为 True, 如果错误, 则为 False; 也可用来判断是否到视频末尾; 再返回一个值, 为每一帧的图像, 该值是一个三维矩阵; 通用接收方法为: ret,frame = cap.read(); ret: 布尔值; frame: 图像的三维矩阵; 这样 ret 存储布尔值, frame 存储图像; 若使用一个变量来接收两个值, 如: frame = cap.read() 则 frame 为一个元组, 原来使用 frame 处需更改为 frame[1] """ while cap.isOpened(): ret_flag, img_camera = cap.read() cv2.imshow("camera", img_camera) # 每帧数据延时 1ms, 延时为0, 读取的是静态帧 k = cv2.waitKey(1) # 按下 's' 保存截图 if k == ord('s'): cv2.imwrite("test.jpg", img_camera) # 按下 'q' 退出 if k == ord('q'): break # 释放所有摄像头 cap.release() # 删除建立的所有窗口 cv2.destroyAllWindows()
python3 how_to_use_camera.py
提取头像
下面的视频自己可以随便找一部,人脸多的。代码的目的就是截取人,为之后收集人脸做铺垫。以下代码中有存在路径,自己注意修改自己文件夹目录在哪。
文件名:capture.py
import cv2 print(cv2.__version__) # /home/zjipc/Documents vc = cv2.VideoCapture('D:\picture\mda-ihuh3xjpbi20te2c.MP4' ) # 读入视频文件 c = 1 caps = 0 if vc.isOpened(): # 判断是否正常打开 rval, frame = vc.read() caps = vc.get(7) # 获取视频总帧数 else: rval = False timeF = int(caps / 30) timeF = 100 # 视频帧计数间隔频率 i = 1 while rval: # 循环读取视频帧 rval, frame = vc.read() # print("c="+str(c)) if (c % timeF == 0): # 每隔timeF帧进行存储操作 path = 'D:\picture\pi' + str(i).zfill(6) + '.jpg' print("正在保存:" + path) cv2.imwrite(path, frame) # 存储为图像 i = i + 1 c = c + 1 cv2.waitKey(1) vc.release()
python3 capture.py
人脸收集
可以通俗的理解为注册人脸
文件名:get_faces_from_camera.py
import dlib # 人脸处理的库 Dlib import numpy as np # 数据处理的库 Numpy import cv2 # 图像处理的库 OpenCv import os # 读写文件 import shutil # 读写文件 # Dlib 正向人脸检测器 detector = dlib.get_frontal_face_detector() # Dlib 68 点特征预测器 predictor = dlib.shape_predictor('data/data_dlib/shape_predictor_68_face_landmarks.dat') # OpenCv 调用摄像头 cap = cv2.VideoCapture(0) # 设置视频参数 cap.set(3, 480) # 人脸截图的计数器 cnt_ss = 0 # 存储人脸的文件夹 current_face_dir = 0 # 保存 photos/csv 的路径 path_photos_from_camera = "data/data_faces_from_camera/" path_csv_from_photos = "data/data_csvs_from_camera/" # 新建保存人脸图像文件和数据CSV文件夹 def pre_work_mkdir(): # 新建文件夹 if os.path.isdir(path_photos_from_camera): pass else: os.mkdir(path_photos_from_camera) if os.path.isdir(path_csv_from_photos): pass else: os.mkdir(path_csv_from_photos) pre_work_mkdir() ##### optional/可选, 默认关闭 ##### # 删除之前存的人脸数据文件夹 def pre_work_deldir(): # 删除之前存的人脸数据文件夹 # 删除 "/data_faces_from_camera/person_x/"... folders_rd = os.listdir(path_photos_from_camera) for i in range(len(folders_rd)): shutil.rmtree(path_photos_from_camera+folders_rd[i]) csv_rd = os.listdir(path_csv_from_photos) for i in range(len(csv_rd)): os.remove(path_csv_from_photos+csv_rd[i]) # 如果有之前录入的人脸 # 在之前 person_x 的序号按照 person_x+1 开始录入 if os.listdir("data/data_faces_from_camera/"): # 获取已录入的最后一个人脸序号 person_list = os.listdir("data/data_faces_from_camera/") person_list.sort() person_num_latest = int(str(person_list[-1]).split("_")[-1]) person_cnt = person_num_latest # 如果第一次存储或者没有之前录入的人脸, 按照 person_1 开始录入 else: person_cnt = 0 # 之后用来控制是否保存图像的 flag save_flag = 1 while cap.isOpened(): # 480 height * 640 width flag, img_rd = cap.read() kk = cv2.waitKey(1) img_gray = cv2.cvtColor(img_rd, cv2.COLOR_RGB2GRAY) # 人脸数 faces faces = detector(img_gray, 0) # 待会要写的字体 font = cv2.FONT_HERSHEY_COMPLEX # 按下 'n' 新建存储人脸的文件夹 if kk == ord('n'): person_cnt += 1 current_face_dir = path_photos_from_camera + "person_" + str(person_cnt) os.makedirs(current_face_dir) print('\n') print("新建的人脸文件夹: ", current_face_dir) # 将人脸计数器清零 cnt_ss = 0 # 检测到人脸 if len(faces) != 0: # 矩形框 for k, d in enumerate(faces): # 计算矩形大小 # (x,y), (宽度width, 高度height) pos_start = tuple([d.left(), d.top()]) pos_end = tuple([d.right(), d.bottom()]) # 计算矩形框大小 height = (d.bottom() - d.top()) width = (d.right() - d.left()) hh = int(height/2) ww = int(width/2) # 设置颜色 / The color of rectangle of faces detected color_rectangle = (255, 255, 255) # if (d.right()+ww) > 640 or (d.bottom()+hh > 480) or (d.left()-ww < 0) or (d.top()-hh < 0): # cv2.putText(img_rd, "OUT OF RANGE", (20, 300), font, 0.8, (0, 0, 255), 1, cv2.LINE_AA) # color_rectangle = (0, 0, 255) # save_flag = 0 # else: # color_rectangle = (255, 255, 255) # save_flag = 1 cv2.rectangle(img_rd, tuple([d.left() - ww, d.top() - hh]), tuple([d.right() + ww, d.bottom() + hh]), color_rectangle, 2) # 根据人脸大小生成空的图像 im_blank = np.zeros((int(height*2), width*2, 3), np.uint8) if save_flag: # 按下 's' 保存摄像头中的人脸到本地 if kk == ord('s'): if os.path.isdir(current_face_dir): cnt_ss += 1 # for ii in range(height*2): # for jj in range(width*2): # im_blank[ii][jj] = img_rd[d.top()-hh + ii][d.left()-ww + jj] # print(im_blank[ii][jj]) cv2.imwrite(current_face_dir + "/img_face_" + str(cnt_ss) + ".jpg", img_rd) print("写入本地:", str(current_face_dir) + "/img_face_" + str(cnt_ss) + ".jpg") else: print("请在按 'S' 之前先按 'N' 来建文件夹 / Please press 'N' before 'S'") # 显示人脸数 cv2.putText(img_rd, "Faces: " + str(len(faces)), (20, 100), font, 0.8, (0, 255, 0), 1, cv2.LINE_AA) # 添加说明 cv2.putText(img_rd, "Face Register", (20, 40), font, 1, (0, 0, 0), 1, cv2.LINE_AA) cv2.putText(img_rd, "N: New face folder", (20, 350), font, 0.8, (0, 0, 0), 1, cv2.LINE_AA) cv2.putText(img_rd, "S: Save current face", (20, 400), font, 0.8, (0, 0, 0), 1, cv2.LINE_AA) cv2.putText(img_rd, "Q: Quit", (20, 450), font, 0.8, (0, 0, 0), 1, cv2.LINE_AA) # 按下 'q' 键退出 if kk == ord('q'): break # 窗口显示 # cv2.namedWindow("camera", 0) # 如果需要摄像头窗口大小可调 cv2.imshow("camera", img_rd) # 释放摄像头 cap.release() # 删除建立的窗口 cv2.destroyAllWindows()
python3 get_faces_from_camera.py
接着上面的代码写入后,在data文件里,会自动生成照片管理文件,这也是在ubuntu下做的好处,我们可以借助库建立数据存储,不需要在安装数据库。
人脸识别的核心
文件名:face_reco_from_camera.py
#coding=utf-8 # 摄像头实时人脸识别 import dlib # 人脸处理的库 Dlib import numpy as np # 数据处理的库 numpy import cv2 # 图像处理的库 OpenCv import pandas as pd # 数据处理的库 Pandas # 人脸识别模型,提取128D的特征矢量 # face recognition model, the object maps human faces into 128D vectors facerec = dlib.face_recognition_model_v1("data/data_dlib/dlib_face_recognition_resnet_model_v1.dat") # 计算两个128D向量间的欧式距离 def return_euclidean_distance(feature_1, feature_2): feature_1 = np.array(feature_1) feature_2 = np.array(feature_2) np.linalg.norm(feature_1 - feature_2) dist = np.sqrt(np.sum(np.square(feature_1 - feature_2))) print("e_distance: ", dist) if dist > 0.4: return "diff" else: return "same" # 处理存放所有人脸特征的 csv path_features_known_csv = "data/features_all.csv" csv_rd = pd.read_csv(path_features_known_csv, header=None) # 存储的特征人脸个数 # print(csv_rd.shape[0]) # 用来存放所有录入人脸特征的数组 features_known_arr = [] # 读取已知人脸数据 # known faces for i in range(csv_rd.shape[0]): features_someone_arr = [] print(csv_rd.ix[i, :]) for j in range(0, len(csv_rd.ix[i, :])): features_someone_arr.append(csv_rd.ix[i, :][j]) features_known_arr.append(features_someone_arr) print("Faces in Database:", len(features_known_arr)) # Dlib 检测器和预测器 detector = dlib.get_frontal_face_detector() predictor = dlib.shape_predictor('data/data_dlib/shape_predictor_68_face_landmarks.dat') # 创建 cv2 摄像头对象 cap = cv2.VideoCapture(0) # cap.set(propId, value) # 设置视频参数,propId 设置的视频参数,value 设置的参数值 cap.set(3, 480) # 返回一张图像多张人脸的 128D 特征 def get_128d_features(img_gray): faces = detector(img_gray, 1) if len(faces) != 0: face_des = [] for i in range(len(faces)): shape = predictor(img_gray, faces[i]) face_des.append(facerec.compute_face_descriptor(img_gray, shape)) else: face_des = [] return face_des # cap.isOpened() 返回 true/false 检查初始化是否成功 while cap.isOpened(): flag, img_rd = cap.read() kk = cv2.waitKey(1) # 取灰度 img_gray = cv2.cvtColor(img_rd, cv2.COLOR_RGB2GRAY) # 人脸数 faces faces = detector(img_gray, 0) # 待会要写的字体 font = cv2.FONT_HERSHEY_COMPLEX # 存储当前摄像头中捕获到的所有人脸的坐标/名字 pos_namelist = [] name_namelist = [] # 按下 q 键退出 if kk == ord('q'): break else: # 检测到人脸 if len(faces) != 0: # 获取当前捕获到的图像的所有人脸的特征,存储到 features_cap_arr features_cap_arr = [] for i in range(len(faces)): shape = predictor(img_rd, faces[i]) features_cap_arr.append(facerec.compute_face_descriptor(img_rd, shape)) # 遍历捕获到的图像中所有的人脸 for k in range(len(faces)): # 让人名跟随在矩形框的下方 # 确定人名的位置坐标 # 先默认所有人不认识,是 unknown name_namelist.append("unknown") # 每个捕获人脸的名字坐标 pos_namelist.append(tuple([faces[k].left(), int(faces[k].bottom() + (faces[k].bottom() - faces[k].top())/4)])) # 对于某张人脸,遍历所有存储的人脸特征 for i in range(len(features_known_arr)): print("with person_", str(i+1), "the") # 将某张人脸与存储的所有人脸数据进行比对 compare = return_euclidean_distance(features_cap_arr[k], features_known_arr[i]) if compare == "same": # 找到了相似脸 # 在这里修改 person_1, person_2 ... 的名字 # 这里只写了前三个 # 可以在这里改称 Jack, Tom and others # Here you can modify the names shown on the camera if i == 0: name_namelist[k] = "zhangsan" elif i == 1: name_namelist[k] = "jiaqing" elif i == 2: name_namelist[k] = "Unkonw" # 矩形框 for kk, d in enumerate(faces): # 绘制矩形框 cv2.rectangle(img_rd, tuple([d.left(), d.top()]), tuple([d.right(), d.bottom()]), (0, 255, 255), 2) # 在人脸框下面写人脸名字 for i in range(len(faces)): cv2.putText(img_rd, name_namelist[i], pos_namelist[i], font, 0.8, (0, 255, 255), 1, cv2.LINE_AA) print("Name list now:", name_namelist, "\n") # 窗口显示 cv2.imshow("camera", img_rd) # 释放摄像头 cap.release() # 删除建立的窗口 cv2.destroyAllWindows()
python3 face_reco_from_camera.py
这两张照片都是我自己,为什么第二张识别不了呢? 这就是深度学习中的问题了。机器它识别也有限的。所以我们要训练模型,使它的准确度更高。当然,手机美颜照片也是不行的
训练
文件名:get_features_into_CSV.py
# 从人脸图像文件中提取人脸特征存入 CSV # Get features from images and save into features_all.csv # 增加录入多张人脸到 CSV 的功能 # return_128d_features() 获取某张图像的 128D 特征 # write_into_csv() 获取某个路径下所有图像的特征,并写入 CSV # compute_the_mean() 从 CSV 中读取 128D 特征,并计算特征均值 import cv2 import os import dlib from skimage import io import csv import numpy as np import pandas as pd #import io # 要读取人脸图像文件的路径 path_photos_from_camera = "data/data_faces_from_camera/" # 储存人脸特征 csv 的路径 path_csv_from_photos = "data/data_csvs_from_camera/" # Dlib 正向人脸检测器 detector = dlib.get_frontal_face_detector() # Dlib 人脸预测器 predictor = dlib.shape_predictor("data/data_dlib/shape_predictor_5_face_landmarks.dat") # Dlib 人脸识别模型 # Face recognition model, the object maps human faces into 128D vectors facerec = dlib.face_recognition_model_v1("data/data_dlib/dlib_face_recognition_resnet_model_v1.dat") # 返回单张图像的 128D 特征 def return_128d_features(path_img): img = io.imread(path_img) img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) faces = detector(img_gray, 1) print("检测到人脸的图像:", path_img, "\n") # 因为有可能截下来的人脸再去检测,检测不出来人脸了 # 所以要确保是 检测到人脸的人脸图像 拿去算特征 if len(faces) != 0: shape = predictor(img_gray, faces[0]) face_descriptor = facerec.compute_face_descriptor(img_gray, shape) else: face_descriptor = 0 print("no face") # print(face_descriptor) return face_descriptor # 将文件夹中照片特征提取出来, 写入 CSV # path_faces_personX: 图像文件夹的路径 # path_csv_from_photos: 要生成的 CSV 路径 def write_into_csv(path_faces_personX, path_csv_from_photos): photos_list = os.listdir(path_faces_personX) with open(path_csv_from_photos, "w", newline="") as csvfile: writer = csv.writer(csvfile) if photos_list: for i in range(len(photos_list)): # 调用return_128d_features()得到128d特征 print("正在读的人脸图像:", path_faces_personX + "/" + photos_list[i]) features_128d = return_128d_features(path_faces_personX + "/" + photos_list[i]) print(features_128d) # 遇到没有检测出人脸的图片跳过 if features_128d == 0: i += 1 else: writer.writerow(features_128d) else: print("Warning: Empty photos in "+path_faces_personX+'/') writer.writerow("") # 读取某人所有的人脸图像的数据,写入 person_X.csv faces = os.listdir(path_photos_from_camera) faces.sort() for person in faces: print("##### " + person + " #####") print(path_csv_from_photos + person + ".csv") write_into_csv(path_photos_from_camera + person, path_csv_from_photos + person + ".csv") print('\n') # 从 CSV 中读取数据,计算 128D 特征的均值 def compute_the_mean(path_csv_from_photos): column_names = [] # 128D 特征 for feature_num in range(128): column_names.append("features_" + str(feature_num + 1)) # 利用 pandas 读取 csv rd = pd.read_csv(path_csv_from_photos, names=column_names) if rd.size != 0: # 存放 128D 特征的均值 feature_mean_list = [] for feature_num in range(128): tmp_arr = rd["features_" + str(feature_num + 1)] tmp_arr = np.array(tmp_arr) # 计算某一个特征的均值 tmp_mean = np.mean(tmp_arr) feature_mean_list.append(tmp_mean) else: feature_mean_list = [] return feature_mean_list # 存放所有特征均值的 CSV 的路径 path_csv_from_photos_feature_all = "data/features_all.csv" # 存放人脸特征的 CSV 的路径 path_csv_from_photos = "data/data_csvs_from_camera/" with open(path_csv_from_photos_feature_all, "w", newline="") as csvfile: writer = csv.writer(csvfile) csv_rd = os.listdir(path_csv_from_photos) csv_rd.sort() print("##### 得到的特征均值 / The generated average values of features stored in: #####") for i in range(len(csv_rd)): feature_mean_list = compute_the_mean(path_csv_from_photos + csv_rd[i]) print(path_csv_from_photos + csv_rd[i]) writer.writerow(feature_mean_list)
python3 get_features_into_CSV.py
这就基本完了,祝你们毕业快乐!
Copyright © 2003-2013 www.wpsshop.cn 版权所有,并保留所有权利。