caffe-ubuntu1604-gtx850m-i7-4710hq----VGG_ILSVRC_16_layers.caffemodel

 

 c++调用vgg16: 

./build/install/bin/classification    \
 /media/whale/wsWin10/wsCaffe/model-zoo/VGG16//deploy.prototxt  \
  /media/whale/wsWin10/wsCaffe/model-zoo/VGG16/VGG_ILSVRC_16_layers.caffemodel \
   data/ilsvrc12/imagenet_mean.binaryproto \
   /media/whale/wsWin10/wsCaffe/model-zoo/VGG16/synset_words.txt \
   /media/whale/wsWin10/images/person/2.jpg

 然后就报错了。

然后

whale@sea:/media/whale/wsWin10/wsUbuntu16.04/DlFrames/caffe$ ./build/install/bin/classification     /media/whale/wsWin10/wsCaffe/model-zoo/VGG16//deploy.prototxt    /media/whale/wsWin10/wsCaffe/model-zoo/VGG16/VGG_ILSVRC_16_layers.caffemodel    data/ilsvrc12/imagenet_mean.binaryproto    ./3labels.txt    /media/whale/wsWin10/images/person/2.jpg
labels_.size() = 3 output_layer->channels()  = 3 ---------- Prediction for /media/whale/wsWin10/images/person/2.jpg ----------
0.3333 - "1000"
0.3333 - "2000"
0.3333 - "3000"
whale@sea:/media/whale/wsWin10/wsUbuntu16.04/DlFrames/caffe$ ./build/install/bin/classification     /media/whale/wsWin10/wsCaffe/model-zoo/VGG16//deploy.prototxt    /media/whale/wsWin10/wsCaffe/model-zoo/VGG16/VGG_ILSVRC_16_layers.caffemodel    data/ilsvrc12/imagenet_mean.binaryproto    ./3labels.txt    /media/whale/wsWin10/images/person/3.jpg
labels_.size() = 3 output_layer->channels()  = 3 ---------- Prediction for /media/whale/wsWin10/images/person/3.jpg ----------
0.3333 - "1000"
0.3333 - "2000"
0.3333 - "3000"

只能给3个标签，不然就报错。然后，。。。，这个模型是假的吗？

还是什么是假的？

 

keras-python调用vgg16: 

python-keras接口调用模型要简单些，只需要标签文件，和keras模型就可以了。

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
# get_ipython().magic(u'matplotlib inline')


# ### 显示图像

# In[2]:

img_path = './data/elephant.jpg'
img_path = '/media/whale/wsWin10/images/dog/0c02094a98d126cf541c4318188699a5.jpg'
img_path = '/media/whale/wsWin10/images/dog/dd92db98b99479db3619f62c724757a4.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）

# In[3]:

model = VGG16(include_top=True, weights='imagenet')
print(" type(model) =  ",   type(model))


# In[4]:

x = image.img_to_array(img)

x = np.expand_dims(x, axis=0)

x = preprocess_input(x)

print( "x.max()  = ",   x.max())

scores = model.predict(x)


# In[10]:

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) = ",  np.argmax(scores))
print('result is ', lines[np.argmax(scores)])
class_table.close()

import sys
sys.exit()

 

。。。/wsWin10/wsPycharm/sklearnStu/Keras-Tutorials/08.vgg-16.py
Using TensorFlow backend.
2018-01-16 17:35:28.541700: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX AVX2 FMA
2018-01-16 17:35:28.627059: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:892] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2018-01-16 17:35:28.627317: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1030] Found device 0 with properties: 
name: GeForce GTX 850M major: 5 minor: 0 memoryClockRate(GHz): 0.9015
pciBusID: 0000:01:00.0
totalMemory: 3.95GiB freeMemory: 3.63GiB
2018-01-16 17:35:28.627334: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1120] Creating TensorFlow device (/device:GPU:0) -> (device: 0, name: GeForce GTX 850M, pci bus id: 0000:01:00.0, compute capability: 5.0)
(' type(model) =  ', <class 'keras.engine.training.Model'>)
('x.max()  = ', 151.061)
(' scores type: ', <type 'numpy.ndarray'>)
(' scores shape: ', (1, 1000))
('  np.argmax(scores) = ', 235)
('result is ', 'n02106662 German shepherd, German shepherd dog, German police dog, alsatian\n')

Process finished with exit code 0

翻译下： 德国牧羊犬，德国牧羊犬，德国警犬，阿尔萨斯