是否带眼镜检测
作者：互联网
第一个版本：
数据样本，带眼镜的图片2000
不带眼镜的2000张
开始训练模型
# -*- coding: utf-8 -*-
"""
Created on Wed Jul 10 15:39:22 2019
@author: 01
"""
import cv2
import glob
import os
import tensorflow as tf
import numpy as np
import time
tf.reset_default_graph()
#数据集地址
path='C:/Users/01/Desktop/face/data/images/img_small/'
#模型保存地址
model_path='F:/1/model.ckpt'
#将所有的图片resize成100*100
w=100
h=100
c=3
#读取图片
def read_img(path):
cate=[path+x for x in os.listdir(path) if os.path.isdir(path+x)]
imgs=[]
labels=[]
for idx,folder in enumerate(cate):
for im in glob.glob(folder+'/*.jpg'):
print('reading the images:%s'%(im))
img=cv2.imread(im)
img=cv2.resize(img,(w,h))
imgs.append(img)
labels.append(idx)
return np.asarray(imgs,np.float32),np.asarray(labels,np.int32)
data,label=read_img(path)
#打乱顺序
num_example=data.shape[0]
arr=np.arange(num_example)
np.random.shuffle(arr)
data=data[arr]
label=label[arr]
#将所有数据分为训练集和验证集
ratio=0.8
s=np.int(num_example*ratio)
x_train=data[:s]
y_train=label[:s]
x_val=data[s:]
y_val=label[s:]
#-----------------构建网络----------------------
#占位符
x=tf.placeholder(tf.float32,shape=[None,w,h,c],name='x')
y_=tf.placeholder(tf.int32,shape=[None,],name='y_')
def inference(input_tensor, train, regularizer):
with tf.variable_scope('layer1-conv1'):
conv1_weights = tf.get_variable("weight",[5,5,3,32],initializer=tf.truncated_normal_initializer(stddev=0.1))
conv1_biases = tf.get_variable("bias", [32], initializer=tf.constant_initializer(0.0))
conv1 = tf.nn.conv2d(input_tensor, conv1_weights, strides=[1, 1, 1, 1], padding='SAME')
relu1 = tf.nn.relu(tf.nn.bias_add(conv1, conv1_biases))
with tf.name_scope("layer2-pool1"):
pool1 = tf.nn.max_pool(relu1, ksize = [1,2,2,1],strides=[1,2,2,1],padding="VALID")
with tf.variable_scope("layer3-conv2"):
conv2_weights = tf.get_variable("weight",[5,5,32,64],initializer=tf.truncated_normal_initializer(stddev=0.1))
conv2_biases = tf.get_variable("bias", [64], initializer=tf.constant_initializer(0.0))
conv2 = tf.nn.conv2d(pool1, conv2_weights, strides=[1, 1, 1, 1], padding='SAME')
relu2 = tf.nn.relu(tf.nn.bias_add(conv2, conv2_biases))
with tf.name_scope("layer4-pool2"):
pool2 = tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
with tf.variable_scope("layer5-conv3"):
conv3_weights = tf.get_variable("weight",[3,3,64,128],initializer=tf.truncated_normal_initializer(stddev=0.1))
conv3_biases = tf.get_variable("bias", [128], initializer=tf.constant_initializer(0.0))
conv3 = tf.nn.conv2d(pool2, conv3_weights, strides=[1, 1, 1, 1], padding='SAME')
relu3 = tf.nn.relu(tf.nn.bias_add(conv3, conv3_biases))
with tf.name_scope("layer6-pool3"):
pool3 = tf.nn.max_pool(relu3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
with tf.variable_scope("layer7-conv4"):
conv4_weights = tf.get_variable("weight",[3,3,128,128],initializer=tf.truncated_normal_initializer(stddev=0.1))
conv4_biases = tf.get_variable("bias", [128], initializer=tf.constant_initializer(0.0))
conv4 = tf.nn.conv2d(pool3, conv4_weights, strides=[1, 1, 1, 1], padding='SAME')
relu4 = tf.nn.relu(tf.nn.bias_add(conv4, conv4_biases))
with tf.name_scope("layer8-pool4"):
pool4 = tf.nn.max_pool(relu4, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
nodes = 6*6*128
reshaped = tf.reshape(pool4,[-1,nodes])
with tf.variable_scope('layer9-fc1'):
fc1_weights = tf.get_variable("weight", [nodes, 1024],
initializer=tf.truncated_normal_initializer(stddev=0.1))
if regularizer != None: tf.add_to_collection('losses', regularizer(fc1_weights))
fc1_biases = tf.get_variable("bias", [1024], initializer=tf.constant_initializer(0.1))
fc1 = tf.nn.relu(tf.matmul(reshaped, fc1_weights) + fc1_biases)
if train: fc1 = tf.nn.dropout(fc1, 0.5)
with tf.variable_scope('layer10-fc2'):
fc2_weights = tf.get_variable("weight", [1024, 512],
initializer=tf.truncated_normal_initializer(stddev=0.1))
if regularizer != None: tf.add_to_collection('losses', regularizer(fc2_weights))
fc2_biases = tf.get_variable("bias", [512], initializer=tf.constant_initializer(0.1))
fc2 = tf.nn.relu(tf.matmul(fc1, fc2_weights) + fc2_biases)
if train: fc2 = tf.nn.dropout(fc2, 0.5)
with tf.variable_scope('layer11-fc3'):
fc3_weights = tf.get_variable("weight", [512, 2],
initializer=tf.truncated_normal_initializer(stddev=0.1))
if regularizer != None: tf.add_to_collection('losses', regularizer(fc3_weights))
fc3_biases = tf.get_variable("bias", [2], initializer=tf.constant_initializer(0.1))
logit = tf.matmul(fc2, fc3_weights) + fc3_biases
return logit
#---------------------------网络结束---------------------------
regularizer = tf.contrib.layers.l2_regularizer(0.0001)
logits = inference(x,False,regularizer)
#(小处理)将logits乘以1赋值给logits_eval，定义name，方便在后续调用模型时通过tensor名字调用输出tensor
b = tf.constant(value=1,dtype=tf.float32)
logits_eval = tf.multiply(logits,b,name='logits_eval')
loss=tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits, labels=y_)
train_op=tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)
correct_prediction = tf.equal(tf.cast(tf.argmax(logits,1),tf.int32), y_)
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]
#训练和测试数据，可将n_epoch设置更大一些
n_epoch=10
batch_size=64
saver=tf.train.Saver()
sess=tf.Session()
sess.run(tf.global_variables_initializer())
for epoch in range(n_epoch):
start_time = time.time()
#training
train_loss, train_acc, n_batch = 0, 0, 0
for x_train_a, y_train_a in minibatches(x_train, y_train, batch_size, shuffle=True):
_,err,ac=sess.run([train_op,loss,acc], feed_dict={x: x_train_a, y_: y_train_a})
train_loss += err; train_acc += ac; n_batch += 1
print(" train loss: %f" % (np.sum(train_loss)/ n_batch))
print(" train acc: %f" % (np.sum(train_acc)/ n_batch))
#validation
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})
val_loss += err; val_acc += ac; n_batch += 1
print(" validation loss: %f" % (np.sum(val_loss)/ n_batch))
print(" validation acc: %f" % (np.sum(val_acc)/ n_batch))
saver.save(sess,model_path)
sess.close()
大概10次后，训练准确率达到99%，验证准确率80%左右，
测试模型
# -*- coding: utf-8 -*-
import cv2
import tensorflow as tf
import numpy as np
path1 = "./1.jpg"
path2 = "./2.jpg"
path3 = "./3.jpg"
path4 = "./4.jpg"
path5 = "./face5.jpg"
path6 = "./face6.jpg"
path7 = "./face7.jpg"
path8 = "./face8.jpg"
face_dict = {1:'Has Glass',0:'No Glass'}
w=100
h=100
c=3
def read_one_image(path):
img = cv2.imread(path)
print(img)
img = cv2.resize(img,(w,h))
return np.asarray(img)
with tf.Session() as sess:
data = []
data1 = read_one_image(path1)
data2 = read_one_image(path2)
data3 = read_one_image(path3)
data4 = read_one_image(path4)
data5 = read_one_image(path5)
data6 = read_one_image(path6)
data7 = read_one_image(path7)
data8 = read_one_image(path8)
data.append(data1)
data.append(data2)
data.append(data3)
data.append(data4)
data.append(data5)
data.append(data6)
data.append(data7)
data.append(data8)
saver = tf.train.import_meta_graph('./models/model.ckpt.meta')
saver.restore(sess,tf.train.latest_checkpoint('./models/'))
graph = tf.get_default_graph()
x = graph.get_tensor_by_name("x:0")
feed_dict = {x:data}
logits = graph.get_tensor_by_name("logits_eval:0")
classification_result = sess.run(logits,feed_dict)
#打印出预测矩阵
print(classification_result)
#打印出预测矩阵每一行最大值的索引
#print(tf.argmax(classification_result,1).eval())
#根据索引通过字典对应人脸的分类
output = []
output = tf.argmax(classification_result,1).eval()
for i in range(len(output)):
print("No.",i+1,"face is belong to:"+face_dict[output[i]])
如果想要c++调用模型，需要把ckpt的模型文件转换为pb的形式，代码如下
# -*- coding: utf-8 -*-
"""
Created on Mon Jul 15 08:48:53 2019
@author: 01
"""
import cv2
import tensorflow as tf
import numpy as np
def freeze_graph(input_checkpoint,output_graph):
'''
:param input_checkpoint:
:param output_graph: PB模型保存路径
:return:
'''
# checkpoint = tf.train.get_checkpoint_state(model_folder) #检查目录下ckpt文件状态是否可用
# input_checkpoint = checkpoint.model_checkpoint_path #得ckpt文件路径
# 指定输出的节点名称,该节点名称必须是原模型中存在的节点
output_node_names = "logits_eval"
saver = tf.train.import_meta_graph(input_checkpoint + '.meta', clear_devices=True)
graph = tf.get_default_graph() # 获得默认的图
input_graph_def = graph.as_graph_def() # 返回一个序列化的图代表当前的图
with tf.Session() as sess:
saver.restore(sess, input_checkpoint) #恢复图并得到数据
output_graph_def = tf.graph_util.convert_variables_to_constants( # 模型持久化，将变量值固定
sess=sess,
input_graph_def=input_graph_def,# 等于:sess.graph_def
output_node_names=output_node_names.split(","))# 如果有多个输出节点，以逗号隔开
with tf.gfile.GFile(output_graph, "wb") as f: #保存模型
f.write(output_graph_def.SerializeToString()) #序列化输出
print("%d ops in the final graph." % len(output_graph_def.node)) #得到当前图有几个操作节点
# for op in graph.get_operations():
# print(op.name, op.values())
# 输入ckpt模型路径
input_checkpoint='C:/Users/01/Desktop/face/CNN_Face_Glass_Classfy-master/model/model.ckpt'
# 输出pb模型的路径
out_pb_path="models/pb/frozen_model.pb"
# 调用freeze_graph将ckpt转为pb
freeze_graph(input_checkpoint,out_pb_path)
然后就是c++调用pb文件。
#include<opencv2\opencv.hpp>
#include<opencv2\dnn.hpp>
#include<stdlib.h>
#include<iostream>
#include <Windows.h>
#include<string.h>
#include <cstdlib>
#include "lib\facedetect-dll.h"
#include "lib\facedetect-dll.h"
using namespace cv;
using namespace std;
using namespace dnn;
String labels_txt_file = "F:/123.txt";
#define DETECT_BUFFER_SIZE 0x20000
int Glasses_ide(const char* path)
{
cout << "==================眼镜检测开始==========================" << endl;
//=======对图像进行头像裁剪================
//const char* path = "E:/fangtu1/CNN_Face_Glass_Classfy-master/face9.jpg";
//const char* path = "E:/fangtu1/1.png";
//face_dec(path);
//Pr_Fo(path);
//(path);
//=======对图像进行头像裁剪================
unsigned char * pBuffer = (unsigned char *)malloc(DETECT_BUFFER_SIZE);
Mat src = imread(path);//如果预测结果出错，请修改一下图片路径，为什么？不知道
//检测人脸位置
Mat gray;
cvtColor(src, gray, CV_BGR2GRAY);
auto pResults = facedetect_multiview_reinforce(pBuffer, (unsigned char*)(gray.ptr(0)), gray.cols, gray.rows, (int)gray.step, 1.2f, 2, 48, 0, 1);
Mat m_roi_1(7, 7, CV_32FC2, Scalar(1, 3));
for (int i = 0; i < (pResults ? *pResults : 0); i++)
{
//cout << "==================眼镜检测开始==========================" << endl;
short * p = ((short*)(pResults + 1)) + 142 * i;
int x = p[0];
int y = p[1];
int w = p[2];
int h = p[3];
int neighbors = p[4];
int angle = p[5];
//const char *saveFilePath = "F:\\00res.jpg";
//printf("face_rect=[%d, %d, %d, %d], neighbors=%d, angle=%d\n", x, y, w, h, neighbors, angle);
//cv::rectangle(src, Rect(x-w/2, y, w, h), Scalar(0, 255, 0), 2);
//cv::rectangle(src, Rect(x, y, w, h), Scalar(0, 255, 0), 2);
//cv::rectangle(src, Rect(x + w + w / 16, y + h / 3, w / 5, h / 5), Scalar(255, 0, 0), 2);
//cv::rectangle(src, Rect(x + w + w / 16, y + 4 * h / 3, w / 5, h / 5), Scalar(0, 0, 225), 2);
//cout << "hehe" << x << y << w << h << endl;
// save roi image
//设定感兴趣的区域ROI
Mat m_roi_1 = src(cv::Rect(x, y, w, h));
//imshow("Image Classfic11ation1", m_roi_1);
imwrite("hear.jpg", m_roi_1);
}
Mat src1 = imread("hear.jpg");
//检测人脸位置
if (src1.empty())//判断是否为空，判断图片是否导入
{
cout << "error:no img" << endl;
}
String weights = "F:/output/frozen_model.pb";
Net net = readNetFromTensorflow(weights);
DWORD timestart = GetTickCount(); //定义事件节点，测试运行效率
if (net.empty())//判断模型是否调起。
{
cout << "error :no model" << endl;
}
vector <String> labels = readClassNames(); //读取分类类别
Mat rgb;
int ww = 100;//图像的长和宽
int hh = 100;
resize(src1, src, Size(ww, hh));
cvtColor(src, rgb, COLOR_BGR2RGB);
//cout << "src" << src<< endl;
Mat inputBlob = blobFromImage(src, 0.00390625f, Size(ww, hh), Scalar(), false, false);//false如果用caffe训练的要改成true,tensordlow的不用改变
Mat prob;
net.setInput(inputBlob, "x");//x是输入的名字，必须和tensorflow定义的名字一致。
prob = net.forward("logits_eval");//logits_eval是输出的名字，必须和tensorflow定义的名字一致。
DWORD timeend = GetTickCount();
Mat probMat = prob.reshape(1, 1);
//cout << "pro" << probMat << endl;
Point classNumber;
double classProb;
minMaxLoc(probMat, NULL, &classProb, NULL, &classNumber);
int classidx = classNumber.x;
printf("\n current image classification : %s, possible : %.2f\n", labels.at(classidx).c_str(), classProb);
putText(src, labels.at(classidx), Point(20, 20), FONT_HERSHEY_SIMPLEX, 1.0, Scalar(0, 0, 255), 2, 8);
cout << "用时(毫秒):" << timeend - timestart << endl;
// 显示文本
cout << "==================眼镜检测结束==========================" << endl;
cout << endl;
cout << endl;
imshow("Image Classfication", src);
system("del hear.jpg");
waitKey(1);
//free(pBuffer);
system("pause");
return 0;
}
vector <String>readClassNames()
{
vector <String>classNames;
fstream fp(labels_txt_file);
if (!fp.is_open())
{
cout << "does not open" << endl;
exit(-1);
}
string name;
while (!fp.eof())
{
getline(fp, name);
if (name.length())
classNames.push_back(name);
}
fp.close();
return classNames;
}
虽然可以使用但是准确度不高，改进版本
1：定位人脸，把人脸裁剪出来作为训练数据。代码如下
# created at 2018-01-22
# updated at 2018-09-29
# Author: coneypo
# Blog: http://www.cnblogs.com/AdaminXie
# GitHub: https://github.com/coneypo/Dlib_face_cut
import dlib # 人脸识别的库dlib
import numpy as np # 数据处理的库numpy
import cv2 # 图像处理的库OpenCv
import os
# 读取图像的路径
path_read = "./data/images/faces_for_test/1/1/"
# 用来存储生成的单张人脸的路径
path_save = "./data/images/3/"
# Delete old images
def clear_images():
imgs = os.listdir(path_save)
for img in imgs:
os.remove(path_save + img)
print("clean finish", '\n')
clear_images()
# Dlib 预测器
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('data/dlib/shape_predictor_68_face_landmarks.dat')
# Dlib 检测
jj=1
for i in os.listdir(path_read):
path_1read=path_read+i
print(path_1read)
img = cv2.imread(path_1read)
a=img.shape
jj=jj+1
faces = detector(img, 1)
print("人脸数：", len(faces), '\n')
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()])
# 计算矩形框大小
height = d.bottom()-d.top()
if a[1]>=d.right():
pos_end = tuple([d.right(), d.bottom()])
width = d.right()-d.left()
else:
pos_end = tuple([a[1], d.bottom()])
width =a[1]-d.left()
# 根据人脸大小生成空的图像
try:
img_blank = np.zeros((height, width, 3), np.uint8)
for i in range(height):
for j in range(width):
img_blank[i][j] = img[d.top()+i][d.left()+j]
# cv2.imshow("face_"+str(k+1), img_blank)
# 存在本地
print("Save to:", path_save+"img_face_"+str(k+1)+".jpg")
cv2.imwrite(path_save+"img_face3_"+str(k+1)+"_"+str(jj)+".jpg", img_blank)
except:
print("发生异常")
else:
print("没有异常")
再次训练，测试后准确率达到90%左右，还是达不到要求。
再次改进
直接定位到眼镜区域，把其作为训练样本代码如下
# created at 2018-01-22
# updated at 2018-09-29
# Author: coneypo
# Blog: http://www.cnblogs.com/AdaminXie
# GitHub: https://github.com/coneypo/Dlib_face_cut
import dlib # 人脸识别的库dlib
import numpy as np # 数据处理的库numpy
import cv2 # 图像处理的库OpenCv
import os
# 读取图像的路径
path_read = "./data/images/faces_for_test/1/1/"
# 用来存储生成的单张人脸的路径
path_save = "./data/images/img_small/0/"
# Delete old images
def clear_images():
imgs = os.listdir(path_save)
for img in imgs:
os.remove(path_save + img)
print("clean finish", '\n')
clear_images()
# Dlib 预测器
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor('data/dlib/shape_predictor_68_face_landmarks.dat')
# Dlib 检测
jj=1
for i in os.listdir(path_read):
path_1read=path_read+i
print(path_1read)
img = cv2.imread(path_1read)
a=img.shape
jj=jj+1
faces = detector(img, 1)
print("人脸数：", len(faces), '\n')
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()])
# 计算矩形框大小
height = d.bottom()-d.top()
if a[1]>=d.right():
pos_end = tuple([d.right(), d.bottom()])
width = d.right()-d.left()
else:
pos_end = tuple([a[1], d.bottom()])
width =a[1]-d.left()
# 根据人脸大小生成空的图像
try:
b=height/2
img_blank = np.zeros((int ((4*b)/5), width, 3), np.uint8)
for i in range(int((4*b)/5)):
for j in range(width):
img_blank[i][j] = img[d.top()+int((b)/5)+i][d.left()+j]
# cv2.imshow("face_"+str(k+1), img_blank)
# 存在本地
print("Save to:", path_save+"img_face_"+str(k+1)+".jpg")
cv2.imwrite(path_save+"img_face3_"+str(k+1)+"_"+str(jj)+".jpg", img_blank)
except:
print("发生异常")
else:
print("没有异常")
再次测试，准确度达到99%左右，nice
标签：img,检测,眼镜,initializer,train,graph,是否,tf,path 来源： https://blog.csdn.net/zhouguangfei0717/article/details/97921162