我基本上去了 CNTK 网站并提取了一些代码来获取图像的特征向量（最后一层之前的一层）。
这就是我现在所拥有的，它是一个 iPython Notebook
# importing library
import os,sys
import numpy as np
import cntk as C
from cntk import load_model, combine
import cntk.io.transforms as xforms
from cntk.logging import graph
from cntk.logging.graph import get_node_outputs
import zipfile
import shutil
try:
from urllib.request import urlretrieve
except ImportError:
from urllib import urlretrieve
def download_grocery_data():
if not os.path.exists(os.path.join("Grocery", "testImages")):
filename = os.path.join("Grocery.zip")
if not os.path.exists(filename):
url = "https://www.cntk.ai/DataSets/Grocery/Grocery.zip"
print('Downloading data from ' + url + '...')
urlretrieve(url, filename)
try:
print('Extracting ' + filename + '...')
with zipfile.ZipFile(filename) as myzip:
myzip.extractall()
# if platform != "win32":
testfile = os.path.join("Grocery", "test.txt")
unixfile = os.path.join("Grocery", "test_unix.txt")
out = open(unixfile, 'w')
with open(testfile) as f:
for line in f:
out.write(line.replace('\\', '/'))
out.close()
shutil.move(unixfile, testfile)
finally:
os.remove(filename)
print('Done.')
else:
print('Data already available at ' + '/Grocery')
download_grocery_data()
try:
from urllib.request import urlretrieve
except ImportError:
from urllib import urlretrieve
# Add models here like this: (category, model_name, model_url)
models = (('Image Classification', 'AlexNet_ImageNet_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/AlexNet_ImageNet_CNTK.model'),
('Image Classification', 'AlexNet_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/AlexNet_ImageNet_Caffe.model'),
('Image Classification', 'InceptionV3_ImageNet_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/InceptionV3_ImageNet_CNTK.model'),
('Image Classification', 'BNInception_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/BNInception_ImageNet_Caffe.model'),
('Image Classification', 'ResNet18_ImageNet_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/ResNet18_ImageNet_CNTK.model'),
('Image Classification', 'ResNet34_ImageNet_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/ResNet34_ImageNet_CNTK.model'),
('Image Classification', 'ResNet50_ImageNet_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/ResNet50_ImageNet_CNTK.model'),
('Image Classification', 'ResNet20_CIFAR10_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/ResNet20_CIFAR10_CNTK.model'),
('Image Classification', 'ResNet110_CIFAR10_CNTK', 'https://www.cntk.ai/Models/CNTK_Pretrained/ResNet110_CIFAR10_CNTK.model'),
('Image Classification', 'ResNet50_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/ResNet50_ImageNet_Caffe.model'),
('Image Classification', 'ResNet101_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/ResNet101_ImageNet_Caffe.model'),
('Image Classification', 'ResNet152_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/ResNet152_ImageNet_Caffe.model'),
('Image Classification', 'VGG16_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/VGG16_ImageNet_Caffe.model'),
('Image Classification', 'VGG19_ImageNet_Caffe', 'https://www.cntk.ai/Models/Caffe_Converted/VGG19_ImageNet_Caffe.model'),
('Image Object Detection', 'Fast-RCNN_grocery100', 'https://www.cntk.ai/Models/FRCN_Grocery/Fast-RCNN_grocery100.model'),
('Image Object Detection', 'Fast-RCNN_Pascal', 'https://www.cntk.ai/Models/FRCN_Pascal/Fast-RCNN.model'))
def download_model(model_file_name, model_url):
# model_dir = os.path.dirname(os.path.abspath(__file__))
# filename = os.path.join(model_dir, model_file_name)
filename = os.path.join(model_file_name)
if not os.path.exists(filename):
print('Downloading model from ' + model_url + ', may take a while...')
urlretrieve(model_url, filename)
print('Saved model as ' + filename)
else:
print('CNTK model already available at ' + filename)
def download_model_by_name(model_name):
if model_name.endswith('.model'):
model_name = model_name[:-6]
model = next((x for x in models if x[1]==model_name), None)
if model is None:
print("ERROR: Unknown model name '%s'." % model_name)
list_available_models()
else:
download_model(model_name + '.model', model[2])
def list_available_models():
print("\nAvailable models (for more information see Readme.md):")
max_cat = max(len(x[1]) for x in models)
max_name = max(len(x[1]) for x in models)
print("{:<{width}} {}".format('Model name', 'Category', width=max_name))
print("{:-<{width}} {:-<{width_cat}}".format('', '', width=max_name, width_cat=max_cat))
for model in sorted(models):
print("{:<{width}} {}".format(model[1], model[0], width=max_name))
sys.path.append(os.path.join( "..", "DataSets", "Grocery"))
# from install_grocery import download_grocery_data
download_grocery_data()
sys.path.append(os.path.join("..", "..", "..", "PretrainedModels"))
# from download_model import download_model_by_name
download_model_by_name("ResNet18_ImageNet_CNTK")
def create_mb_source(image_height, image_width, num_channels, map_file):
transforms = [xforms.scale(width=image_width, height=image_height, channels=num_channels, interpolations='linear')]
return C.io.MinibatchSource(
C.io.ImageDeserializer(map_file, C.io.StreamDefs(
features=C.io.StreamDef(field='image', transforms=transforms),
labels=C.io.StreamDef(field='label', shape=1000))),
randomize=False)
def eval_and_write(model_file, node_name, output_file, minibatch_source, num_objects):
# load model and pick desired node as output
loaded_model = load_model(model_file)
node_in_graph = loaded_model.find_by_name(node_name)
output_nodes = combine([node_in_graph.owner])
# evaluate model and get desired node output
print("Evaluating model for output node %s" % node_name)
features_si = minibatch_source['features']
with open(output_file, 'wb') as results_file:
for i in range(0, num_objects):
mb = minibatch_source.next_minibatch(1)
output = output_nodes.eval(mb[features_si])
# write results to file
out_values = output[0].flatten()
np.savetxt(results_file, out_values[np.newaxis], fmt="%.6f")
def main():
# define location of model and data and check existence
model_file = os.path.join("ResNet18_ImageNet_CNTK.model")
print(model_file)
map_file = os.path.join("Grocery", "test.txt")
print(map_file)
# os.chdir(os.path.join("..", "DataSets", "Grocery"))
if not (os.path.exists(model_file)):
print("model bhetena")
if not (os.path.exists(map_file)):
print("map file bhetena")
if not (os.path.exists(model_file) and os.path.exists(map_file)):
print("Please run 'python install_data_and_model.py' first to get the required data and model.")
exit(0)
# create minibatch source
image_height = 224
image_width = 224
num_channels = 3
minibatch_source = create_mb_source(image_height, image_width, num_channels, map_file)
# use this to print all node names of the model (and knowledge of the model to pick the correct one)
# node_outputs = get_node_outputs(load_model(model_file))
# for out in node_outputs: print("{0} {1}".format(out.name, out.shape))
# use this to get 1000 class predictions (not yet softmaxed!)
# node_name = "z"
# output_file = os.path.join(base_folder, "predOutput.txt")
# output_file = os.path.join("predOutput.txt")
# use this to get 512 features from the last but one layer of ResNet_18
node_name = "z.x"
# output_file = os.path.join(base_folder, "layerOutput.txt")
output_file = os.path.join("layerOutput.txt")
# evaluate model and write out the desired layer output
eval_and_write(model_file, node_name, output_file, minibatch_source, num_objects=5)
print("Done. Wrote output to %s" % output_file)
main()
此时它的作用是下载模型，下载测试图像，但模型无法读取图像。
这是我得到的错误，
ResNet18_ImageNet_CNTK.model
Grocery\test.txt
Evaluating model for output node z.x
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-5-283fca1e6e4a> in <module>()
80 print("Done. Wrote output to %s" % output_file)
81
---> 82 main()
<ipython-input-5-283fca1e6e4a> in main()
76
77 # evaluate model and write out the desired layer output
---> 78 eval_and_write(model_file, node_name, output_file, minibatch_source, num_objects=5)
79
80 print("Done. Wrote output to %s" % output_file)
<ipython-input-5-283fca1e6e4a> in eval_and_write(model_file, node_name, output_file, minibatch_source, num_objects)
18 with open(output_file, 'wb') as results_file:
19 for i in range(0, num_objects):
---> 20 mb = minibatch_source.next_minibatch(1)
21 output = output_nodes.eval(mb[features_si])
22
c:\users\t540p\appdata\local\programs\python\python35\lib\site-packages\cntk\internal\swig_helper.py in wrapper(*args, **kwds)
67 @wraps(f)
68 def wrapper(*args, **kwds):
---> 69 result = f(*args, **kwds)
70 map_if_possible(result)
71 return result
c:\users\t540p\appdata\local\programs\python\python35\lib\site-packages\cntk\io\__init__.py in next_minibatch(self, minibatch_size_in_samples, input_map, device, num_data_partitions, partition_index)
327 minibatch_size_in_samples,
328 num_data_partitions,
--> 329 partition_index, device)
330
331 if not mb:
c:\users\t540p\appdata\local\programs\python\python35\lib\site-packages\cntk\cntk_py.py in get_next_minibatch(self, *args)
2967
2968 def get_next_minibatch(self, *args):
-> 2969 return _cntk_py.MinibatchSource_get_next_minibatch(self, *args)
2970 MinibatchSource_swigregister = _cntk_py.MinibatchSource_swigregister
2971 MinibatchSource_swigregister(MinibatchSource)
RuntimeError: Cannot open file 'testImages/WIN_20160803_11_28_42_Pro.jpg'
[CALL STACK]
> Microsoft::MSR::CNTK::IDataReader:: SupportsDistributedMBRead
- CreateCompositeDataReader (x3)
- vcomp_fork (x3)
- vcomp_atomic_div_r8
- vcomp_fork
- CreateCompositeDataReader (x4)
- CNTK::Dictionary:: ~Dictionary (x3)
这是错误的屏幕截图，
