08.VGG16调用
程序说明
时间：2016年11月16日
说明：该程序是一个包含两个隐藏层的神经网络。
Very Deep Convolutional Networks for Large-Scale Image Recognition
K. Simonyan, A. Zisserman
arXiv:1409.1556
1.加载keras模块
from keras.applications.vgg16 import VGG16
from keras.preprocessing import image
from keras.applications.vgg16 import preprocess_input
from keras.models import Model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
Using TensorFlow backend.
显示图像
img_path = './data/elephant.jpg'
img = image.load_img(img_path, target_size=(224, 224))
plt.imshow(img)
plt.show()
加载VGG16模型（包含全连接层）
model = VGG16(include_top=True, weights='imagenet')
print(type(model))
<class 'keras.engine.training.Model'>
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
print(x.max())
scores = model.predict(x)
151.061
class_table = open('./data/synset_words', 'r')
lines = class_table.readlines()
print("scores type: ", type(scores))
print("scores shape: ", scores.shape)
print(np.argmax(scores))
print('result is ', lines[np.argmax(scores)])
class_table.close()
('scores type: ', <type 'numpy.ndarray'>)
('scores shape: ', (1, 1000))
386
('result is ', 'n02504458 African elephant, Loxodonta africana\n')
del model
加载VGG16模型（不包含包含全连接层）
model = VGG16(weights='imagenet', include_top=False)
x = image.img_to_array(img)
x = np.expand_dims(x, axis=0)
x = preprocess_input(x)
features = model.predict(x)
print(features.shape)
(1, 7, 7, 512)
model.summary()
____________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
====================================================================================================
input_2 (InputLayer) (None, None, None, 3) 0
____________________________________________________________________________________________________
block1_conv1 (Convolution2D) (None, None, None, 64)1792 input_2[0][0]
____________________________________________________________________________________________________
block1_conv2 (Convolution2D) (None, None, None, 64)36928 block1_conv1[0][0]
____________________________________________________________________________________________________
block1_pool (MaxPooling2D) (None, None, None, 64)0 block1_conv2[0][0]
____________________________________________________________________________________________________
block2_conv1 (Convolution2D) (None, None, None, 12873856 block1_pool[0][0]
____________________________________________________________________________________________________
block2_conv2 (Convolution2D) (None, None, None, 128147584 block2_conv1[0][0]
____________________________________________________________________________________________________
block2_pool (MaxPooling2D) (None, None, None, 1280 block2_conv2[0][0]
____________________________________________________________________________________________________
block3_conv1 (Convolution2D) (None, None, None, 256295168 block2_pool[0][0]
____________________________________________________________________________________________________
block3_conv2 (Convolution2D) (None, None, None, 256590080 block3_conv1[0][0]
____________________________________________________________________________________________________
block3_conv3 (Convolution2D) (None, None, None, 256590080 block3_conv2[0][0]
____________________________________________________________________________________________________
block3_pool (MaxPooling2D) (None, None, None, 2560 block3_conv3[0][0]
____________________________________________________________________________________________________
block4_conv1 (Convolution2D) (None, None, None, 5121180160 block3_pool[0][0]
____________________________________________________________________________________________________
block4_conv2 (Convolution2D) (None, None, None, 5122359808 block4_conv1[0][0]
____________________________________________________________________________________________________
block4_conv3 (Convolution2D) (None, None, None, 5122359808 block4_conv2[0][0]
____________________________________________________________________________________________________
block4_pool (MaxPooling2D) (None, None, None, 5120 block4_conv3[0][0]
____________________________________________________________________________________________________
block5_conv1 (Convolution2D) (None, None, None, 5122359808 block4_pool[0][0]
____________________________________________________________________________________________________
block5_conv2 (Convolution2D) (None, None, None, 5122359808 block5_conv1[0][0]
____________________________________________________________________________________________________
block5_conv3 (Convolution2D) (None, None, None, 5122359808 block5_conv2[0][0]
____________________________________________________________________________________________________
block5_pool (MaxPooling2D) (None, None, None, 5120 block5_conv3[0][0]
====================================================================================================
Total params: 14714688
____________________________________________________________________________________________________
提取block5_pool层特征
model_extractfeatures = Model(input=model.input, output=model.get_layer('block5_pool').output)
block5_pool_features = model_extractfeatures.predict(x)
print(type(block5_pool_features))
print(block5_pool_features.shape)
<type 'numpy.ndarray'>
(1, 7, 7, 512)
feature_image = block5_pool_features[:,:,:,0].reshape(7,7)
plt.imshow(feature_image)
plt.show()
