这段时间正在学习tensorflow的卷积神经网络部分,为了对卷积神经网络能够有一个更深的了解,自己动手实现一个例程是比较好的方式,所以就选了一个这样比较有点意思的项目。
项目的github地址:github 喜欢的话就给个Star吧。
想要她认得我,就需要给她一些我的照片,让她记住我的人脸特征,为了让她区分我和其他人,还需要给她一些其他人的照片做参照,所以就需要两组数据集来让她学习,如果想让她多认识几个人,那多给她几组图片集学习就可以了。下面就开始让我们来搭建这个能认识我的"她"。
运行环境
下面为软件的运行搭建系统环境。
系统: window或linux
软件: python 3.x 、 tensorflow
python支持库:
tensorflow:
pip install tensorflow #cpu版本 pip install rensorflow-gpu #gpu版本,需要cuda与cudnn的支持,不清楚的可以选择cpu版
numpy:
pip install numpy
opencv:
pip install opencv-python
dlib:
pip install dlib
获取本人图片集
获取本人照片的方式当然是拍照了,我们需要通过程序来给自己拍照,如果你自己有照片,也可以用那些现成的照片,但前提是你的照片足够多。这次用到的照片数是10000张,程序运行后,得坐在电脑面前不停得给自己的脸摆各种姿势,这样可以提高训练后识别自己的成功率,在程序中加入了随机改变对比度与亮度的模块,也是为了提高照片样本的多样性。
程序中使用的是dlib来识别人脸部分,也可以使用opencv来识别人脸,在实际使用过程中,dlib的识别效果比opencv的好,但opencv识别的速度会快很多,获取10000张人脸照片的情况下,dlib大约花费了1小时,而opencv的花费时间大概只有20分钟。opencv可能会识别一些奇怪的部分,所以综合考虑之后我使用了dlib来识别人脸。
get_my_faces.py
import cv2 import dlib import os import sys import random output_dir = './my_faces' size = 64 if not os.path.exists(output_dir): os.makedirs(output_dir) # 改变图片的亮度与对比度 def relight(img, light=1, bias=0): w = img.shape[1] h = img.shape[0] #image = [] for i in range(0,w): for j in range(0,h): for c in range(3): tmp = int(img[j,i,c]*light + bias) if tmp > 255: tmp = 255 elif tmp < 0: tmp = 0 img[j,i,c] = tmp return img #使用dlib自带的frontal_face_detector作为我们的特征提取器 detector = dlib.get_frontal_face_detector() # 打开摄像头 参数为输入流,可以为摄像头或视频文件 camera = cv2.VideoCapture(0) index = 1 while True: if (index <= 10000): print('Being processed picture %s' % index) # 从摄像头读取照片 success, img = camera.read() # 转为灰度图片 gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 使用detector进行人脸检测 dets = detector(gray_img, 1) for i, d in enumerate(dets): x1 = d.top() if d.top() > 0 else 0 y1 = d.bottom() if d.bottom() > 0 else 0 x2 = d.left() if d.left() > 0 else 0 y2 = d.right() if d.right() > 0 else 0 face = img[x1:y1,x2:y2] # 调整图片的对比度与亮度, 对比度与亮度值都取随机数,这样能增加样本的多样性 face = relight(face, random.uniform(0.5, 1.5), random.randint(-50, 50)) face = cv2.resize(face, (size,size)) cv2.imshow('image', face) cv2.imwrite(output_dir+'/'+str(index)+'.jpg', face) index += 1 key = cv2.waitKey(30) & 0xff if key == 27: break else: print('Finished!') break
在这里我也给出一个opencv来识别人脸的代码示例:
import cv2 import os import sys import random out_dir = './my_faces' if not os.path.exists(out_dir): os.makedirs(out_dir) # 改变亮度与对比度 def relight(img, alpha=1, bias=0): w = img.shape[1] h = img.shape[0] #image = [] for i in range(0,w): for j in range(0,h): for c in range(3): tmp = int(img[j,i,c]*alpha + bias) if tmp > 255: tmp = 255 elif tmp < 0: tmp = 0 img[j,i,c] = tmp return img # 获取分类器 haar = cv2.CascadeClassifier('haarcascade_frontalface_default.xml') # 打开摄像头 参数为输入流,可以为摄像头或视频文件 camera = cv2.VideoCapture(0) n = 1 while 1: if (n <= 10000): print('It`s processing %s image.' % n) # 读帧 success, img = camera.read() gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) faces = haar.detectMultiScale(gray_img, 1.3, 5) for f_x, f_y, f_w, f_h in faces: face = img[f_y:f_y+f_h, f_x:f_x+f_w] face = cv2.resize(face, (64,64)) ''' if n % 3 == 1: face = relight(face, 1, 50) elif n % 3 == 2: face = relight(face, 0.5, 0) ''' face = relight(face, random.uniform(0.5, 1.5), random.randint(-50, 50)) cv2.imshow('img', face) cv2.imwrite(out_dir+'/'+str(n)+'.jpg', face) n+=1 key = cv2.waitKey(30) & 0xff else: break
获取其他人脸图片集
需要收集一个其他人脸的图片集,只要不是自己的人脸都可以,可以在网上找到,这里我给出一个我用到的图片集:
网站地址:http://vis-www.cs.umass.edu/lfw/
图片集下载:http://vis-www.cs.umass.edu/lfw/lfw.tgz
先将下载的图片集,解压到项目目录下的input_img目录下,也可以自己指定目录(修改代码中的input_dir变量)
接下来使用dlib来批量识别图片中的人脸部分,并保存到指定目录下
set_other_people.py
# -*- codeing: utf-8 -*- import sys import os import cv2 import dlib input_dir = './input_img' output_dir = './other_faces' size = 64 if not os.path.exists(output_dir): os.makedirs(output_dir) #使用dlib自带的frontal_face_detector作为我们的特征提取器 detector = dlib.get_frontal_face_detector() index = 1 for (path, dirnames, filenames) in os.walk(input_dir): for filename in filenames: if filename.endswith('.jpg'): print('Being processed picture %s' % index) img_path = path+'/'+filename # 从文件读取图片 img = cv2.imread(img_path) # 转为灰度图片 gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # 使用detector进行人脸检测 dets为返回的结果 dets = detector(gray_img, 1) #使用enumerate 函数遍历序列中的元素以及它们的下标 #下标i即为人脸序号 #left:人脸左边距离图片左边界的距离 ;right:人脸右边距离图片左边界的距离 #top:人脸上边距离图片上边界的距离 ;bottom:人脸下边距离图片上边界的距离 for i, d in enumerate(dets): x1 = d.top() if d.top() > 0 else 0 y1 = d.bottom() if d.bottom() > 0 else 0 x2 = d.left() if d.left() > 0 else 0 y2 = d.right() if d.right() > 0 else 0 # img[y:y+h,x:x+w] face = img[x1:y1,x2:y2] # 调整图片的尺寸 face = cv2.resize(face, (size,size)) cv2.imshow('image',face) # 保存图片 cv2.imwrite(output_dir+'/'+str(index)+'.jpg', face) index += 1 key = cv2.waitKey(30) & 0xff if key == 27: sys.exit(0)
这个项目用到的图片数是10000张左右,如果是自己下载的图片集,控制一下图片的数量避免数量不足,或图片过多带来的内存不够与运行缓慢。
训练模型
有了训练数据之后,通过cnn来训练数据,就可以让她记住我的人脸特征,学习怎么认识我了。
train_faces.py
import tensorflow as tf import cv2 import numpy as np import os import random import sys from sklearn.model_selection import train_test_split my_faces_path = './my_faces' other_faces_path = './other_faces' size = 64 imgs = [] labs = [] def getPaddingSize(img): h, w, _ = img.shape top, bottom, left, right = (0,0,0,0) longest = max(h, w) if w < longest: tmp = longest - w # //表示整除符号 left = tmp // 2 right = tmp - left elif h < longest: tmp = longest - h top = tmp // 2 bottom = tmp - top else: pass return top, bottom, left, right def readData(path , h=size, w=size): for filename in os.listdir(path): if filename.endswith('.jpg'): filename = path + '/' + filename img = cv2.imread(filename) top,bottom,left,right = getPaddingSize(img) # 将图片放大, 扩充图片边缘部分 img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=[0,0,0]) img = cv2.resize(img, (h, w)) imgs.append(img) labs.append(path) readData(my_faces_path) readData(other_faces_path) # 将图片数据与标签转换成数组 imgs = np.array(imgs) labs = np.array([[0,1] if lab == my_faces_path else [1,0] for lab in labs]) # 随机划分测试集与训练集 train_x,test_x,train_y,test_y = train_test_split(imgs, labs, test_size=0.05, random_state=random.randint(0,100)) # 参数:图片数据的总数,图片的高、宽、通道 train_x = train_x.reshape(train_x.shape[0], size, size, 3) test_x = test_x.reshape(test_x.shape[0], size, size, 3) # 将数据转换成小于1的数 train_x = train_x.astype('float32')/255.0 test_x = test_x.astype('float32')/255.0 print('train size:%s, test size:%s' % (len(train_x), len(test_x))) # 图片块,每次取100张图片 batch_size = 100 num_batch = len(train_x) // batch_size x = tf.placeholder(tf.float32, [None, size, size, 3]) y_ = tf.placeholder(tf.float32, [None, 2]) keep_prob_5 = tf.placeholder(tf.float32) keep_prob_75 = tf.placeholder(tf.float32) def weightVariable(shape): init = tf.random_normal(shape, stddev=0.01) return tf.Variable(init) def biasVariable(shape): init = tf.random_normal(shape) return tf.Variable(init) def conv2d(x, W): return tf.nn.conv2d(x, W, strides=[1,1,1,1], padding='SAME') def maxPool(x): return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME') def dropout(x, keep): return tf.nn.dropout(x, keep) def cnnLayer(): # 第一层 W1 = weightVariable([3,3,3,32]) # 卷积核大小(3,3), 输入通道(3), 输出通道(32) b1 = biasVariable([32]) # 卷积 conv1 = tf.nn.relu(conv2d(x, W1) + b1) # 池化 pool1 = maxPool(conv1) # 减少过拟合,随机让某些权重不更新 drop1 = dropout(pool1, keep_prob_5) # 第二层 W2 = weightVariable([3,3,32,64]) b2 = biasVariable([64]) conv2 = tf.nn.relu(conv2d(drop1, W2) + b2) pool2 = maxPool(conv2) drop2 = dropout(pool2, keep_prob_5) # 第三层 W3 = weightVariable([3,3,64,64]) b3 = biasVariable([64]) conv3 = tf.nn.relu(conv2d(drop2, W3) + b3) pool3 = maxPool(conv3) drop3 = dropout(pool3, keep_prob_5) # 全连接层 Wf = weightVariable([8*16*32, 512]) bf = biasVariable([512]) drop3_flat = tf.reshape(drop3, [-1, 8*16*32]) dense = tf.nn.relu(tf.matmul(drop3_flat, Wf) + bf) dropf = dropout(dense, keep_prob_75) # 输出层 Wout = weightVariable([512,2]) bout = weightVariable([2]) #out = tf.matmul(dropf, Wout) + bout out = tf.add(tf.matmul(dropf, Wout), bout) return out def cnnTrain(): out = cnnLayer() cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=out, labels=y_)) train_step = tf.train.AdamOptimizer(0.01).minimize(cross_entropy) # 比较标签是否相等,再求的所有数的平均值,tf.cast(强制转换类型) accuracy = tf.reduce_mean(tf.cast(tf.equal(tf.argmax(out, 1), tf.argmax(y_, 1)), tf.float32)) # 将loss与accuracy保存以供tensorboard使用 tf.summary.scalar('loss', cross_entropy) tf.summary.scalar('accuracy', accuracy) merged_summary_op = tf.summary.merge_all() # 数据保存器的初始化 saver = tf.train.Saver() with tf.Session() as sess: sess.run(tf.global_variables_initializer()) summary_writer = tf.summary.FileWriter('./tmp', graph=tf.get_default_graph()) for n in range(10): # 每次取128(batch_size)张图片 for i in range(num_batch): batch_x = train_x[i*batch_size : (i+1)*batch_size] batch_y = train_y[i*batch_size : (i+1)*batch_size] # 开始训练数据,同时训练三个变量,返回三个数据 _,loss,summary = sess.run([train_step, cross_entropy, merged_summary_op], feed_dict={x:batch_x,y_:batch_y, keep_prob_5:0.5,keep_prob_75:0.75}) summary_writer.add_summary(summary, n*num_batch+i) # 打印损失 print(n*num_batch+i, loss) if (n*num_batch+i) % 100 == 0: # 获取测试数据的准确率 acc = accuracy.eval({x:test_x, y_:test_y, keep_prob_5:1.0, keep_prob_75:1.0}) print(n*num_batch+i, acc) # 准确率大于0.98时保存并退出 if acc > 0.98 and n > 2: saver.save(sess, './train_faces.model', global_step=n*num_batch+i) sys.exit(0) print('accuracy less 0.98, exited!') cnnTrain()
训练之后的数据会保存在当前目录下。
使用模型进行识别
最后就是让她认识我了,很简单,只要运行程序,让摄像头拍到我的脸,她就可以轻松地识别出是不是我了。
is_my_face.py
output = cnnLayer() predict = tf.argmax(output, 1) saver = tf.train.Saver() sess = tf.Session() saver.restore(sess, tf.train.latest_checkpoint('.')) def is_my_face(image): res = sess.run(predict, feed_dict={x: [image/255.0], keep_prob_5:1.0, keep_prob_75: 1.0}) if res[0] == 1: return True else: return False #使用dlib自带的frontal_face_detector作为我们的特征提取器 detector = dlib.get_frontal_face_detector() cam = cv2.VideoCapture(0) while True: _, img = cam.read() gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) dets = detector(gray_image, 1) if not len(dets): #print('Can`t get face.') cv2.imshow('img', img) key = cv2.waitKey(30) & 0xff if key == 27: sys.exit(0) for i, d in enumerate(dets): x1 = d.top() if d.top() > 0 else 0 y1 = d.bottom() if d.bottom() > 0 else 0 x2 = d.left() if d.left() > 0 else 0 y2 = d.right() if d.right() > 0 else 0 face = img[x1:y1,x2:y2] # 调整图片的尺寸 face = cv2.resize(face, (size,size)) print('Is this my face? %s' % is_my_face(face)) cv2.rectangle(img, (x2,x1),(y2,y1), (255,0,0),3) cv2.imshow('image',img) key = cv2.waitKey(30) & 0xff if key == 27: sys.exit(0) sess.close()