本文主要是介绍vgg16-py代码,希望对大家解决编程问题提供一定的参考价值,需要的开发者们随着小编来一起学习吧!
#tensorflow学习笔记(北京大学) vgg16.py 完全解析
#QQ群:476842922(欢迎加群讨论学习
#!/usr/bin/python
#coding:utf-8
import inspect
import os
import numpy as np
import tensorflow as tf
import time
import matplotlib.pyplot as plt
#样本RGB的平均值
VGG_MEAN = [103.939, 116.779, 123.68]
class Vgg16():
def init(self, vgg16_path=None):
if vgg16_path is None:
#返回当前工作目录
vgg16_path = os.path.join(os.getcwd(), “vgg16.npy”)
#遍历其内键值对,导入模型参数
self.data_dict = np.load(vgg16_path, encoding=‘latin1’).item()
def forward(self, images):print("build model started")#获取前向传播开始时间start_time = time.time() #逐个像素乘以255rgb_scaled = images * 255.0 #从GRB转换彩色通道到BRGred, green, blue = tf.split(rgb_scaled,3,3) #减去每个通道的像素平均值,这种操作可以移除图像的平均亮度值#该方法常用在灰度图像上bgr = tf.concat([ blue - VGG_MEAN[0],green - VGG_MEAN[1],red - VGG_MEAN[2]],3)#构建VGG的16层网络(包含5段卷积,3层全连接),并逐层根据命名空间读取网络参数#第一段卷积,含有两个卷积层,后面接最大池化层,用来缩小图片尺寸self.conv1_1 = self.conv_layer(bgr, "conv1_1") #传入命名空间的name,来获取该层的卷积核和偏置,并做卷积运算,最后返回经过激活函数后的值self.conv1_2 = self.conv_layer(self.conv1_1, "conv1_2")#根据传入的pooling名字对该层做相应的池化操作self.pool1 = self.max_pool_2x2(self.conv1_2, "pool1")#第二段卷积,包含两个卷积层,一个最大池化层self.conv2_1 = self.conv_layer(self.pool1, "conv2_1")self.conv2_2 = self.conv_layer(self.conv2_1, "conv2_2")self.pool2 = self.max_pool_2x2(self.conv2_2, "pool2")#第三段卷积,包含三个卷积层,一个最大池化层self.conv3_1 = self.conv_layer(self.pool2, "conv3_1")self.conv3_2 = self.conv_layer(self.conv3_1, "conv3_2")self.conv3_3 = self.conv_layer(self.conv3_2, "conv3_3")self.pool3 = self.max_pool_2x2(self.conv3_3, "pool3")#第四段卷积,包含三个卷积层,一个最大池化层self.conv4_1 = self.conv_layer(self.pool3, "conv4_1")self.conv4_2 = self.conv_layer(self.conv4_1, "conv4_2")self.conv4_3 = self.conv_layer(self.conv4_2, "conv4_3")self.pool4 = self.max_pool_2x2(self.conv4_3, "pool4")#第五段卷积,包含三个卷积层,一个最大池化层self.conv5_1 = self.conv_layer(self.pool4, "conv5_1")self.conv5_2 = self.conv_layer(self.conv5_1, "conv5_2")self.conv5_3 = self.conv_layer(self.conv5_2, "conv5_3")self.pool5 = self.max_pool_2x2(self.conv5_3, "pool5")#第六层全连接#根据命名空间name做加权求和运算self.fc6 = self.fc_layer(self.pool5, "fc6")#经过relu激活函数self.relu6 = tf.nn.relu(self.fc6) #第七层全连接self.fc7 = self.fc_layer(self.relu6, "fc7")self.relu7 = tf.nn.relu(self.fc7)#第八层全连接self.fc8 = self.fc_layer(self.relu7, "fc8")self.prob = tf.nn.softmax(self.fc8, name="prob")#得到全向传播时间end_time = time.time() print(("time consuming: %f" % (end_time-start_time)))#清空本次读取到的模型参数字典self.data_dict = None #定义卷积运算
def conv_layer(self, x, name):#根据命名空间name找到对应卷积层的网络参数with tf.variable_scope(name): #读到该层的卷积核w = self.get_conv_filter(name) #卷积运算conv = tf.nn.conv2d(x, w, [1, 1, 1, 1], padding='SAME') #读到偏置项conv_biases = self.get_bias(name) #加上偏置,并做激活计算result = tf.nn.relu(tf.nn.bias_add(conv, conv_biases)) return result#定义获取卷积核的参数
def get_conv_filter(self, name):#根据命名空间从参数字典中获取对应的卷积核return tf.constant(self.data_dict[name][0], name="filter") #定义获取偏置项的参数
def get_bias(self, name):#根据命名空间从参数字典中获取对应的偏置项return tf.constant(self.data_dict[name][1], name="biases")#定义最大池化操作
def max_pool_2x2(self, x, name):return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME', name=name)#定义全连接层的全向传播操作
def fc_layer(self, x, name):#根据命名空间name做全连接层的计算with tf.variable_scope(name): #获取该层的维度信息列表shape = x.get_shape().as_list() dim = 1for i in shape[1:]:#将每层的维度相乘dim *= i #改变特征图的形状,也就是将得到的多维特征做拉伸操作,只在进入第六层全连接层做该操作x = tf.reshape(x, [-1, dim])#读到权重值w = self.get_fc_weight(name) #读到偏置项值b = self.get_bias(name) #对该层输入做加权求和,再加上偏置result = tf.nn.bias_add(tf.matmul(x, w), b) return result#定义获取权重的函数
def get_fc_weight(self, name): #根据命名空间name从参数字典中获取对应1的权重return tf.constant(self.data_dict[name][0], name="weights")
这篇关于vgg16-py代码的文章就介绍到这儿,希望我们推荐的文章对编程师们有所帮助!