TensorFlow练习23: “恶作剧”

前面有一个帖《OpenCV检测场景内是否有移动物体》我用树莓派做了一个简单的Motion Detection，放在卫生间的，它会在我上大号时自动播放音乐。

我一个人租房，几个盆友周末时常会找我玩，他们觉得我做的Motion Detection很垃圾。于是我就想恶搞一下，用TensorFlow做一个“人脸识别”，在我上大号时播放音乐，如果是别人就播放《张震讲鬼故事》（@xingCI说放屁声更搞）。

我的任务的训练一个模型可以区分“我”和“其它人”的脸。注意，上面“人脸识别”我是加引号的，其实并不是真正的人脸识别，充其量就是个图像分类。如果你要使用真正的人脸识别，可以试试现成的库OpenFace+dlib《使用OpenFace进行人脸识别》。

有人已经把TensorFlow移植到了树莓派，项目地址tensorflow-on-raspberry-pi。

准备数据

本帖需要使用到两组数据：一组是包含我脸的图像，另一组包含其它人人脸的图像。

其它人人脸的收集

找一堆图片，只要不包含自己就行，然后使用OpenCV提取图像中的大脸。

我使用的数据集是前面几帖用到的：《妹子图》《大脸》《imagenet》

提取图像中的人脸，我使用OpenCV，据说使用dlib效果更好。

other_peoples_faces.py：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

import cv2 import os import sys   IMAGE_DIR = '图片目录路径'   OUTPUT_DIR = './other_people' if not os.path . exists ( OUTPUT_DIR ) : os . makedirs ( OUTPUT_DIR )   # http://blog.topspeedsnail.com/archives/10511 # wget https://raw.githubusercontent.com/opencv/opencv/master/data/haarcascades/haarcascade_frontalface_default.xml face_haar = cv2 . CascadeClassifier ( "haarcascade_frontalface_default.xml" )   for ( dirpath , dirnames , filenames ) in os . walk ( IMAGE_DIR ) : for filename in filenames : if filename . endswith ( '.jpg' ) : image_path = os.path . join ( dirpath , filename ) print ( 'process: ' , image_path ) img = cv2 . imread ( image_path ) gray_image = cv2 . cvtColor ( img , cv2 . COLOR_BGR2GRAY ) faces = face_haar . detectMultiScale ( gray_image , 1.3 , 5 ) for face_x , face_y , face_w , face_h in faces : face = img [ face_y : face_y + face_h , face_x : face_x + face_w ]   face = cv2 . resize ( face , ( 64 , 64 ) )   cv2 . imshow ( "img" , face ) cv2 . imwrite ( os.path . join ( OUTPUT_DIR , filename ) , face )   key = cv2 . waitKey ( 30 ) & 0xff if key == 27 : sys . exit ( 0 )

4万多图片，我只提取了1万张脸，应该够使了。

图像大小 64×64

上面是OpenCV做的人脸检测，有了这个数据集又可以反过来训练TensorFlow版本的人脸检测。

斗大熊的脸

给自己拍照1万张，这是我一次拍照最多的一回。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

import cv2 import os import sys   OUTPUT_DIR = './my_faces' if not os.path . exists ( OUTPUT_DIR ) : os . makedirs ( OUTPUT_DIR )   face_haar = cv2 . CascadeClassifier ( "haarcascade_frontalface_default.xml" )   cam = cv2 . VideoCapture ( 0 )   count = 0 while True : print ( count ) if count < 10000 : _ , img = cam . read ( )   gray_image = cv2 . cvtColor ( img , cv2 . COLOR_BGR2GRAY ) faces = face_haar . detectMultiScale ( gray_image , 1.3 , 5 ) for face_x , face_y , face_w , face_h in faces : face = img [ face_y : face_y + face_h , face_x : face_x + face_w ]   face = cv2 . resize ( face , ( 64 , 64 ) ) cv2 . imshow ( 'img' , face ) cv2 . imwrite ( os.path . join ( OUTPUT_DIR , str ( count ) + '.jpg' ) , face ) count += 1 key = cv2 . waitKey ( 30 ) & 0xff if key == 27 : break else : break

在镜头前摇头晃脑、摆pose，戴眼镜、耳机，仰天45，写代码，呲牙咧嘴，玩手机。。。一定要多样化，直到拍1万张大脸。

训练模型

训练数据有了，下面开始训练。

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

import tensorflow as tf import cv2 import numpy as np import os from sklearn . model_selection import train_test_split import random import sys   my_image_path = 'my_faces' others_image_path = 'other_people'   image_data = [ ] label_data = [ ]   def get_padding_size ( image ) : h , w , _ = image . shape longest_edge = max ( h , w ) top , bottom , left , right = ( 0 , 0 , 0 , 0 ) if h < longest_edge : dh = longest_edge - h top = dh / / 2 bottom = dh - top elif w < longest_edge : dw = longest_edge - w left = dw / / 2 right = dw - left else : pass return top , bottom , left , right   def read_data ( img_path , image_h = 64 , image_w = 64 ) : for filename in os . listdir ( img_path ) : if filename . endswith ( '.jpg' ) : filepath = os.path . join ( img_path , filename ) image = cv2 . imread ( filepath )   top , bottom , left , right = get_padding_size ( image ) image_pad = cv2 . copyMakeBorder ( image , top , bottom , left , right , cv2 . BORDER_CONSTANT , value = [ 0 , 0 , 0 ] ) image = cv2 . resize ( image_pad , ( image_h , image_w ) )   image_data . append ( image ) label_data . append ( img_path )   read_data ( others_image_path ) read_data ( my_image_path )   image_data = np . array ( image_data ) label_data = np . array ( [ [ 0 , 1 ] if label == 'my_faces' else [ 1 , 0 ] for label in label_data ] )   train_x , test_x , train_y , test_y = train_test_split ( image_data , label_data , test_size = 0.05 , random_state = random . randint ( 0 , 100 ) )   # image (height=64 width=64 channel=3) train_x = train_x . reshape ( train_x . shape [ 0 ] , 64 , 64 , 3 ) test_x = test_x . reshape ( test_x . shape [ 0 ] , 64 , 64 , 3 )   # nomalize train_x = train_x . astype ( 'float32' ) / 255.0 test_x = test_x . astype ( 'float32' ) / 255.0   print ( len ( train_x ) , len ( train_y ) ) print ( len ( test_x ) , len ( test_y ) )   ############################################################# batch_size = 128 num_batch = len ( train_x ) / / batch _size   X = tf . placeholder ( tf . float32 , [ None , 64 , 64 , 3 ] )    # 图片大小64x64 channel=3 Y = tf . placeholder ( tf . float32 , [ None , 2 ] )   keep_prob_5 = tf . placeholder ( tf . float32 ) keep_prob_75 = tf . placeholder ( tf . float32 )   def panda_joke_cnn ( ) :   W_c1 = tf . Variable ( tf . random_normal ( [ 3 , 3 , 3 , 32 ] , stddev = 0.01 ) ) b_c1 = tf . Variable ( tf . random_normal ( [ 32 ] ) ) conv1 = tf . nn . relu ( tf . nn . bias_add ( tf . nn . conv2d ( X , W_c1 , strides = [ 1 , 1 , 1 , 1 ] , padding = 'SAME' ) , b_c1 ) ) conv1 = tf . nn . max_pool ( conv1 , ksize = [ 1 , 2 , 2 , 1 ] , strides = [ 1 , 2 , 2 , 1 ] , padding = 'SAME' ) conv1 = tf . nn . dropout ( conv1 , keep_prob_5 )   W_c2 = tf . Variable ( tf . random_normal ( [ 3 , 3 , 32 , 64 ] , stddev = 0.01 ) ) b_c2 = tf . Variable ( tf . random_normal ( [ 64 ] ) ) conv2 = tf . nn . relu ( tf . nn . bias_add ( tf . nn . conv2d ( conv1 , W_c2 , strides = [ 1 , 1 , 1 , 1 ] , padding = 'SAME' ) , b_c2 ) ) conv2 = tf . nn . max_pool ( conv2 , ksize = [ 1 , 2 , 2 , 1 ] , strides = [ 1 , 2 , 2 , 1 ] , padding = 'SAME' ) conv2 = tf . nn . dropout ( conv2 , keep_prob_5 )   W_c3 = tf . Variable ( tf . random_normal ( [ 3 , 3 , 64 , 64 ] , stddev = 0.01 ) ) b_c3 = tf . Variable ( tf . random_normal ( [ 64 ] ) ) conv3 = tf . nn . relu ( tf . nn . bias_add ( tf . nn . conv2d ( conv2 , W_c3 , strides = [ 1 , 1 , 1 , 1 ] , padding = 'SAME' ) , b_c3 ) ) conv3 = tf . nn . max_pool ( conv3 , ksize = [ 1 , 2 , 2 , 1 ] , strides = [ 1 , 2 , 2 , 1 ] , padding = 'SAME' ) conv3 = tf . nn . dropout ( conv3 , keep_prob_5 )   # Fully connected layer W_d = tf . Variable ( tf . random_normal ( [ 8 * 16 * 32 , 512 ] , stddev = 0.01 ) ) b_d = tf . Variable ( tf . random_normal ( [ 512 ] ) ) dense = tf . reshape ( conv3 , [ - 1 , W_d . get_shape ( ) . as_list ( ) [ 0 ] ] ) dense = tf . nn . relu ( tf . add ( tf . matmul ( dense , W_d ) , b_d ) ) dense = tf . nn . dropout ( dense , keep_prob_75 )   W_out = tf . Variable ( tf . random_normal ( [ 512 , 2 ] , stddev = 0.01 ) ) b_out = tf . Variable ( tf . random_normal ( [ 2 ] ) ) out = tf . add ( tf . matmul ( dense , W_out ) , b_out ) return out   def train_cnn ( ) : output = panda_joke_cnn ( )   loss = tf . reduce_mean ( tf . nn . softmax_cross_entropy_with_logits ( output , Y ) ) optimizer = tf . train . AdamOptimizer ( learning_rate = 0.001 ) . minimize ( loss )   accuracy = tf . reduce_mean ( tf . cast ( tf . equal ( tf . argmax ( output , 1 ) , tf . argmax ( Y , 1 ) ) , tf . float32 ) )   tf . summary . scalar ( "loss" , loss ) tf . summary . scalar ( "accuracy" , accuracy ) merged_summary_op = tf . summary . merge_all ( )   saver = tf . train . Saver ( ) with tf . Session ( ) as sess : sess . run ( tf . global_variables_initializer ( ) )   summary_writer = tf . summary . FileWriter ( './log' , graph = tf . get_default_graph ( ) )   for e in range ( 50 ) : 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 ] _ , loss_ , summary = sess . run ( [ optimizer , loss , 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 , e * num_batch + i ) print ( e * num_batch + i , loss_ )   if ( e * num_batch + i ) % 100 == 0 : acc = accuracy . eval ( { X : test_x , Y : test_y , keep_prob_5 : 1.0 , keep_prob_75 : 1.0 } ) print ( e * num_batch + i , acc ) # save model if acc > 0.98 : saver . save ( sess , "i_am_a_joke.model" , global_step = e * num_batch + i ) sys . exit ( 0 )   train_cnn ( )

准确率曲线：

下面要做的就是在树莓派上使用模型，代码示例：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34

output = panda_joke_cnn ( ) 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   face_haar = cv2 . CascadeClassifier ( "haarcascade_frontalface_default.xml" ) cam = cv2 . VideoCapture ( 0 )   while True : _ , img = cam . read ( ) gray_image = cv2 . cvtColor ( img , cv2 . COLOR_BGR2GRAY ) faces = face_haar . detectMultiScale ( gray_image , 1.3 , 5 ) for face_x , face_y , face_w , face_h in faces : face = img [ face_y : face_y + face_h , face_x : face_x + face_w ]   face = cv2 . resize ( face , ( 64 , 64 ) )   print ( is_my_face ( face ) )   cv2 . imshow ( 'img' , face ) key = cv2 . waitKey ( 30 ) & 0xff if key == 27 : sys . exit ( 0 )   sess . close ( )

总结：占用内存100多M，准确率还凑合，先用着。

上面方法需要收集两类数据（自己的脸和其他人的脸），有没有办法只使用自己的脸做One-Class分类呢？这样就不需收集其他人的脸了。感觉上可以用autoencoder。

如要转载，请保持本文完整，并注明作者@斗大的熊猫和本文原始地址： http://blog.topspeedsnail.com/archives/10931