摸鱼了快两个月，把之前咕咕咕的补上。
这里就直接贴代码了(有空一定会补上说明的)。
train.py
from skimage import io,transform
import glob
import os
import tensorflow as tf
import numpy as np
import time
#图片路径
#path='D:/face_recognize/texture'
path='D:/faceRecognize/faces'
#定义预处理的图片大小为128*128,3通道
w=128
h=128
c=3
#最外层循环，建立独立训练的列表
for file_num in range(3):
#for file_num in range(1):
print ('testing NO.%d dataset.......' % file_num)
#读取图片方法，包含预处理
def read_img(path):
cate=[path+'/'+x for x in os.listdir(path) if os.path.isdir(path+'/'+x)]
imgs=[]
labs=[]
for idx,folder in enumerate(cate):
for im in glob.glob(folder+'/*.png'):
print('加载图片:%s'%(im))
img=io.imread(im)
img=transform.resize(img,(w,h,c))
imgs.append(img)
labs.append(idx)
return np.asarray(imgs,np.float32),np.asarray(labs,np.int32)
#调用方法进行图片读取，并将数据与标签信息存入两个数组
data,label=read_img(path)
num_example=data.shape[0]
arr=np.arange(num_example)
np.random.shuffle(arr)
#数据数组
data=data[arr]
#label标签数组
label=label[arr]
ratio=0.8
s=np.int(num_example*ratio)
train_x=data[:s]#训练集数据
train_y=label[:s]#训练集标签
#以上是读取训练数据
x_val=data[s:]#测试集数据
y_val=label[s:]#测试集标签
#给x，y_留出占位符，以便未来填充数据,x_img作为深度网络的输入，y_作为网络的标签输出
x=tf.placeholder(tf.float32,shape=[None,w,h,c],name='x')
y_=tf.placeholder(tf.int32,shape=[None,],name='y_')
#dropout的参数
keep_prob = tf.placeholder("float")
#定义卷积网络
def CNNlayer():
#第一个卷积层（128——>64)
conv1=tf.layers.conv2d(
inputs=x,
filters=32,#通道32
kernel_size=[5, 5], #卷积核5*5
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))#步长[1,1]
pool1=tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)#最大池化，卷积核[2*2],步长为2
#第二个卷积层(64->32)
conv2=tf.layers.conv2d(
inputs=pool1,
filters=64,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool2=tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
#第三个卷积层(32->16)
conv3=tf.layers.conv2d(
inputs=pool2,
filters=128,
kernel_size=[3, 3],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool3=tf.layers.max_pooling2d(inputs=conv3, pool_size=[2, 2], strides=2)
#第四个卷积层(16->8)
conv4=tf.layers.conv2d(
inputs=pool3,
filters=128,
kernel_size=[3, 3],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool4=tf.layers.max_pooling2d(inputs=conv4, pool_size=[2, 2], strides=2)
re1 = tf.reshape(pool4, [-1, 8 * 8 * 128])
#全连接层
dense1 = tf.layers.dense(inputs=re1,
units=1024,
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
dense2= tf.layers.dense(inputs=dense1,
units=512,
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
# 设置dropout，否则很容易过拟合
dense3=tf.nn.dropout(dense2, keep_prob)
logits= tf.layers.dense(inputs=dense3,
units=60,#对应输出层个60标签
activation=None,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
return logits
#网络结构定义结束
#调用网络结构函数
logits=CNNlayer()
#选用softmax作为最后激活函数
loss=tf.losses.sparse_softmax_cross_entropy(labels=y_,logits=logits)
# 设置误差代价以交叉熵的形式
train_op=tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
# 用adam的优化算法优化目标函数
correct_prediction = tf.equal(tf.cast(tf.argmax(logits,1),tf.int32), y_)#比较标签数据和网络实际输出数据是否相等，判断准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
acc= tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#定义一个函数，按批次取数据
def minibatches(inputs=None, targets=None, batch_size=None, shuffle=False):
assert len(inputs) == len(targets)
if shuffle:
indices = np.arange(len(inputs))
np.random.shuffle(indices)
for start_idx in range(0, len(inputs) - batch_size + 1, batch_size):
if shuffle:
excerpt = indices[start_idx:start_idx + batch_size]
else:
excerpt = slice(start_idx, start_idx + batch_size)
yield inputs[excerpt], targets[excerpt]
#保存训练好的模型参数
saver=tf.train.Saver(max_to_keep=3)
#min_loss为训练终止标准，也可以直接训练迭代完整个n_epoch
min_loss=1
#定义算子
sess=tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
#开始训练
#迭代n_epoch次，每次抽出的样本数目为batch_size
n_epoch=3000
batch_size=64
for epoch in range(n_epoch):
train_loss, train_acc, n_batch = 0, 0, 0
for x_train_a, y_train_a in minibatches(train_x, train_y, batch_size, shuffle=True):
_,err,ac=sess.run([train_op,loss,acc], feed_dict={x: x_train_a, y_: y_train_a,keep_prob: 0.6})
# 设置dropout的参数为0.6，
if train_loss<min_loss:
min_loss=train_loss
#print("模型误差为"%min_loss)
val_loss, val_acc, n_batch = 0, 0, 0
for x_val_a, y_val_a in minibatches(x_val, y_val, batch_size, shuffle=False):
err, ac = sess.run([loss,acc], feed_dict={x: x_val_a, y_: y_val_a,keep_prob: 1.0})
val_loss += err; val_acc += ac; n_batch += 1
#终止训练的评判标准，也可以训练整个n_epoch，此处没有使用
#if min_loss<1e-6
saver.save(sess,'ckpt1/faces.ckpt',global_step=epoch+1)
print("第%d次训练完成"% file_num)
print ("第%d次测试的准确率"% file_num+"test accuracy： %g"%(val_acc/n_batch))
#这一句结果和上面的相同，都是取平均
#print ("第%d次测试的准确率"% file_num+"test accuracy： %g"%ac.eval(feed_dict={x: x_val_a, y_: y_val_a, keep_prob: 1.0}))
sess.close()test.py
from skimage import io,transform
import glob
import os
import tensorflow as tf
import numpy as np
import time
import cv2
import dlib
import sys
ID=(1511277,1511278,1511279,1511282,1511283,1511286,1511287,1511289,
1511290,1511291,1511292,1511295,1511296,1511298,1511300,1511301,
1511302,1511303,1511304,1511305,1511306,1511307,1511308,1511310,
1511311,1511312,1511313,1511315,1511317,1511318,1511319,1511323,
1511325,1511328,1511329,1511330,1511332,1511333,1511334,1511337,
1511339,1511340,1511341,1511342,1511343,1511344,1511345,1511347,
1511349,1511350,1511351,1511352,1511353,1511355,1511357,1511358,
1511360,1511361,1511363,1511365)
w=128
h=128
c=3
x=tf.placeholder(tf.float32,shape=[None,w,h,c],name='x')
y_=tf.placeholder(tf.int32,shape=[None,],name='y_')
def CNNlayer():
#第一个卷积层（128——>64)
conv1=tf.layers.conv2d(
inputs=x,
filters=32,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool1=tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
#第二个卷积层(64->32)
conv2=tf.layers.conv2d(
inputs=pool1,
filters=64,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool2=tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
#第三个卷积层(32->16)
conv3=tf.layers.conv2d(
inputs=pool2,
filters=128,
kernel_size=[3, 3],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool3=tf.layers.max_pooling2d(inputs=conv3, pool_size=[2, 2], strides=2)
#第四个卷积层(16->8)
conv4=tf.layers.conv2d(
inputs=pool3,
filters=128,
kernel_size=[3, 3],
padding="same",
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01))
pool4=tf.layers.max_pooling2d(inputs=conv4, pool_size=[2, 2], strides=2)
re1 = tf.reshape(pool4, [-1, 8 * 8 * 128])
#全连接层
dense1 = tf.layers.dense(inputs=re1,
units=1024,
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
dense2= tf.layers.dense(inputs=dense1,
units=512,
activation=tf.nn.relu,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
logits= tf.layers.dense(inputs=dense2,
units=60,
activation=None,
kernel_initializer=tf.truncated_normal_initializer(stddev=0.01),
kernel_regularizer=tf.contrib.layers.l2_regularizer(0.003))
return logits
logits=CNNlayer()
predict = tf.argmax(logits, 1)
detector = dlib.get_frontal_face_detector()
saver = tf.train.Saver()
sess = tf.Session()
saver.restore(sess, 'ckpt1/faces.ckpt-9')
user=input("图片（G）还是摄像头（V）:")
if user=="G":
path=input("图片路径名是：")
img = cv2.imread(path)
dets = detector(img, 1)
#print("Number of faces detected: {}".format(len(dets)))
for index, face in enumerate(dets):
print('face {}; left {}; top {}; right {}; bottom {}'.format(index, face.left(), face.top(), face.right(), face.bottom()))
left = face.left()
top = face.top()
right = face.right()
bottom = face.bottom()
cv2.rectangle(img, (left, top), (right, bottom), (0, 255, 0), 3)
io.imsave('temp.png', img)
img1=io.imread('temp.png')
img1=transform.resize(img1,(w,h,c))
cv2.imshow('image',img1)
img1 = img[top:bottom,left:right]
img1=transform.resize(img1,(w,h,c))
# cv2.imshow('image1',img)
res=sess.run(predict, feed_dict={x:[img1]})
print(ID[res[0]])
if len(dets)==0:
img=transform.resize(img,(w,h,c))
res=sess.run(predict, feed_dict={x:[img]})
print(ID[res[0]])
cv2.waitKey(0)
cv2.destroyAllWindows()
cv2.waitKey(0)
cv2.destroyAllWindows()
else:
cap = cv2.VideoCapture(0)
while True:
ret, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
cv2.imshow('frame', frame)
if cv2.waitKey(1) & 0xFF == ord('s'):
cv2.imwrite('image/now.png', frame)
img = cv2.imread("image/now.png")
dets = detector(img, 1)
print("Number of faces detected: {}".format(len(dets)))
for index, face in enumerate(dets):
print('face {}; left {}; top {}; right {}; bottom {}'.format(index,
face.left(), face.top(), face.right(), face.bottom()))
left = face.left()
top = face.top()
right = face.right()
bottom = face.bottom()
img = img[top:bottom,left:right]
img=transform.resize(img,(w,h,c))
res=sess.run(predict, feed_dict={x:[img]})
print(ID[res[0]])
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
7187
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
