GoogLeNet Example

[idata]

Can you please explain the below code line by line, highly appreciate it

dim=(224,224)

EXAMPLES_BASE_DIR='../../'_________________________*_get labels_________________________*_

labels_file=EXAMPLES_BASE_DIR+'data/ilsvrc12/synset_words.txt'

labels=numpy.loadtxt(labels_file,str,delimiter='\t')_________________________*_configure the NCS_________________________*_

mvnc.SetGlobalOption(mvnc.GlobalOption.LOG_LEVEL, 2)

_________________________*_Get a list of ALL the sticks that are plugged in_________________________*_

devices = mvnc.EnumerateDevices()

if len(devices) == 0: print('No devices found') quit()_________________________*_Pick the first stick to run the network_________________________*_

device = mvnc.Device(devices[0])

_________________________*_Open the NCS_________________________*_

device.OpenDevice()

network_blob='graph'

Load blob

with open(network_blob, mode='rb') as f:

blob = f.read()

graph = device.AllocateGraph(blob)

_________________________*_Load the image_________________________*_

ilsvrc_mean = numpy.load(EXAMPLES_BASE_DIR+'data/ilsvrc12/ilsvrc_2012_mean.npy').mean(1).mean(1) #loading the mean file

img = cv2.imread(EXAMPLES_BASE_DIR+'data/images/nps_electric_guitar.png')img=cv2.resize(img,dim)img = img.astype(numpy.float32)img[:,:,0] = (img[:,:,0] - ilsvrc_mean[0])img[:,:,1] = (img[:,:,1] - ilsvrc_mean[1])img[:,:,2] = (img[:,:,2] - ilsvrc_mean[2])_________________________*_Send the image to the NCS_________________________*_

graph.LoadTensor(img.astype(numpy.float16), 'user object')

_________________________*_Get the result from the NCS_________________________*_

output, userobj = graph.GetResult()

_________________________*_Print the results of the inference form the NCS_________________________*_

order = output.argsort()[::-1][:6]

print('\n------- predictions --------')for i in range(0,5): print ('prediction ' + str(i) + ' (probability ' + str(output[order[i]]) + ') is ' + labels[order[i]] + ' label index is: ' + str(order[i]) )_________________________*_Clean up the graph and the device_________________________*_

graph.DeallocateGraph()

device.CloseDevice()

[idata]

@muhammadyou You can find a step-by-step explanation of the typical API workflow @ https://movidius.github.io/blog/ncs-image-classifier/

[idata]

@Tome_at_Intel Can you explain the get labels code? What is the blob code?

[idata]

@Tome_at_Intel Also in Tensorflow example named inception_v1, there is mean and standard deviation values. Do you mind explaining their purpose?

[idata]

@muhammadyou Labels are usually just a string that describes the category within a model.

Regarding the mean, it is calculated from the training data set and is commonly used for mean subtraction. Subtracting the mean will usually create a "centering" effect in the input data.

Afterwards, the input data is usually divided by the standard deviation causing the input data to be scaled down. Subtracting the mean and dividing by the standard deviation is usually known as feature scaling and is a common technique used to accelerate gradient descent convergence.

You can read more about blobs at http://caffe.berkeleyvision.org/tutorial/net_layer_blob.html.