VGGNet论文信息VGGNet是牛津大学视觉几何组(Visual Geometry Group)提出的模型，该模型在2014ImageNet图像分类与定位挑战赛 ILSVRC-2014中取得在分类任务第二，定位任务第一的优异成绩。VGGNet突出的贡献是证明了很小的卷积，通过增加网络深度可以有效提高性能。
VGGNet结构
VGGNet模型有A-E五种结构网络，深度分别为11,11,13,16,19.实际使用的网络包括VGG16和VGG19.本篇文章主要介绍VGG16,并分享VGG16的Keras实现。下图为VGGNet的结构说明。
VGGNet架构图从图中可以看出VGGNet的一个特点就是简单，通过对卷积的堆叠，在卷积之后进行pooling处理构成一个block，共5个block.
通过对上述的分析，VGG16共包含：
13个卷积层，用conv3-xxx表示；
5个池化层，用maxpool表示；
3个全连接层，用FC-xxx表示。
由于VGG16模型中只有13个卷积层和3个全连接层能产生权重参数，故VGG16的16来自于13 3。
VGG的优缺点
优点：
VGG16结构简单，其实VGGNet结构都简单，通过使用3x3大小的卷积核和最大池化层2x2；
通过实验展示了可以通过加深网络来提升模型性能。
缺点：
网络参数多，占用空间大。
Keras实现VGG16
导入包：
代码语言：javascript复制import numpy as np
import warnings
from keras.models import Model
from keras.layers import Flatten
from keras.layers import Dense
from keras.layers import Input
from keras.layers import Conv2D
from keras.layers import MaxPooling2D
from keras.layers import GlobalMaxPooling2D
from keras.layers import GlobalAveragePooling2D
from keras.preprocessing import image
from keras.utils import layer_utils
from keras import backend as K
from keras.applications.imagenet_utils import decode_predictions
from keras.applications.imagenet_utils import preprocess_input
from keras_applications.imagenet_utils import _obtain_input_shape
from keras.engine.topology import get_source_inputs实例化VGG16架构：
代码语言：javascript复制# 实例化VGG16架构
def VGG16(include_top=True, weights='imagenet',
input_tensor=None, input_shape=None,
pooling=None,
classes=1000):
"""
参数：
:param include_top:
是否在网络顶部包含3个全连接层
:param weights:
权重，随机初始化或者使用已在ImageNet上预训练的权重
:param input_tensor:
可选的Keras张量,input_tensor是layers.Input()的输出，
其作为模型的图像输入
:param input_shape:可选的shape tuple
只有在include_top为False,才需要指定input_shape.
输入的形状必须是带有channels_last数据格式如(224,224,3),
或带有channels_first数据格式如(3,224,224).
input_shape必须有3个输入通道，宽度和高度不应小于48.如(200,200,3)就是一个有效值
:param pooling:池化模式
当include_top为False时，
pooling为None表示模型的输出僵尸最后一个卷积层以4D张量输出；
pooling为avg表示全局均值池化将应用于最后一个卷积的输出，即模型的输出将是2D张量；
pooling为max表示将使用全局最大池化。
:param classes:图像分类的类数，可选
只有在include_top为True时才指定，
:return:
"""
if weights not in {'imagenet', None}:
raise ValueError('The `weights` argument should be either '
'`None` (random initialization) or `imagenet` '
'(pre-training on ImageNet).')
if weights == 'imagenet' and include_top and classes != 1000:
raise ValueError('If using `weights` as imagenet with `include_top`'
' as true, `classes` should be 1000')
# 确定正确的输入形状
input_shape = _obtain_input_shape(input_shape,
default_size=224,
min_size=48,
data_format=K.image_data_format(),
require_flatten=include_top)
if input_tensor is None:
img_input = Input(shape=input_shape)
else:
if not K.is_keras_tensor(input_tensor):
img_input = Input(tensor=input_tensor, shape=input_shape)
else:
img_input = input_tensor
# Block 1
x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv1')(img_input)
x = Conv2D(64, (3, 3), activation='relu', padding='same', name='block1_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
# Block 2
x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv1')(x)
x = Conv2D(128, (3, 3), activation='relu', padding='same', name='block2_conv2')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
# Block 3
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv1')(x)
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv2')(x)
x = Conv2D(256, (3, 3), activation='relu', padding='same', name='block3_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
# Block 4
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv1')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv2')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block4_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
# Block 5
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv1')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv2')(x)
x = Conv2D(512, (3, 3), activation='relu', padding='same', name='block5_conv3')(x)
x = MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
#
if include_top:
# 分类 block
x = Flatten(name='flatten')(x)
x = Dense(4096, activation='relu', name='fc1')(x)
x = Dense(4096, activation='relu', name='fc2')(x)
x = Dense(classes, activation='softmax', name='predictions')(x)
else:
# 当 include_top为False时，设置pooling
if pooling == 'avg':
x = GlobalAveragePooling2D()(x)
elif pooling == 'max':
x = GlobalMaxPooling2D()(x)
# 确保模型考虑了所有的input_tensor
if input_tensor is not None:
inputs = get_source_inputs(input_tensor)
else:
inputs = img_input
# 创建模型.
model = Model(inputs, x, name='vgg16')
# 加载权重
if weights == 'imagenet':
if include_top:
weights_path ="vgg16_weights_tf_dim_ordering_tf_kernels.h5"
#weights_path = get_file('vgg16_weights_tf_dim_ordering_tf_kernels.h5',WEIGHTS_PATH,cache_subdir='model')
else:
weights_path = "vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5"
#weights_path = get_file('vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5',WEIGHTS_PATH_NO_TOP,cache_subdir='model')
model.load_weights(weights_path)
if K.backend() == 'theano':
layer_utils.convert_all_kernels_in_model(model)
if K.image_data_format() == 'channels_first':
if include_top:
maxpool = model.get_layer(name='block5_pool')
shape = maxpool.output_shape[1:]
dense = model.get_layer(name='fc1')
layer_utils.convert_dense_weights_data_format(dense, shape, 'channels_first')
if K.backend() == 'tensorflow':
warnings.warn('You are using the TensorFlow backend, yet you '
'are using the Theano '
'image data format convention '
'(`image_data_format="channels_first"`). '
'For best performance, set '
'`image_data_format="channels_last"` in '
'your Keras config '
'at ~/.keras/keras.json.')
return model定义绘图工具：
代码语言：javascript复制import matplotlib.pyplot as plt
def percent(value):
return '%.2f%%' % (value * 100)
def show_plot(preds):
# 取得前五個最可能的類別及機率
pred = decode_predictions(preds, top=5)[0]
# 整理预测结果,value
values = []
bar_label = []
for element in pred:
values.append(element[2])
bar_label.append(element[1])
# 绘图并保存
fig = plt.figure(u"Top-5 预测结果")
ax = fig.add_subplot(111)
ax.bar(range(len(values)), values, tick_label=bar_label, width=0.5, fc='g')
ax.set_ylabel(u'probability')
ax.set_title(u'Top-5')
for a, b in zip(range(len(values)), values):
ax.text(a, b 0.0005, percent(b), ha='center', va='bottom', fontsize=7)
fig = plt.gcf()
plt.show()测试：
代码语言：javascript复制model = VGG16(include_top=True, weights='imagenet')
img_path = 'cat.jpg'
# 图片尺寸变化shape->(224,224,3)
img = image.load_img(img_path, target_size=(224, 224))
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print('输入图片大小:', x.shape)
# 预测
preds = model.predict(x)
print('预测值:', decode_predictions(preds))输入图片：cat.jpg
cat预测结果
预测结果如上图可看出cat的概率是最大的。
模型下载：
https://github.com/fchollet/deep-learning-models/releases/download/v0.1/vgg16_weights_tf_dim_ordering_tf_kernels.h5
https://github.com/fchollet/deep-learning-models/releases/download/v0.1/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5
参考: https://github.com/handsomeboy/vgg16
https://github.com/fchollet/deep-learning-models/blob/master/vgg16.py
技术分享：机器学习学研社
更多技术,请关注公众号
