

In [1]:

    
import caffe
caffe.set_mode_cpu()
import numpy as np
import matplotlib.pyplot as plt
plt.rcParams['image.cmap'] = 'gray'
%matplotlib inline
from keras.datasets import mnist









    



Using TensorFlow backend.

Preprocess dataset.



In [2]:

    
data = mnist.load_data()[1][0]
# Normalize data.
data = data / data.max()
plt.imshow(data[0, :, :])









    Out[2]:





<matplotlib.image.AxesImage at 0x7fcfbf6daac8>

Reshape the data so it fits into the caffe net.



In [3]:

    
seven = data[0, :, :]
print(seven.shape)
seven = seven[np.newaxis, ...]
print(seven.shape)









    



(28, 28)
(1, 28, 28)

In Caffe the inference model is often seperate from the model for training and suffixed wit _deploy. It ususally has an Input layer that can be freely filled with data, instead of a Data layer with a fixed dataset.

# %load example_caffe_mnist_model_deploy.prototxt name: "ExampleCaffeMNISTModelDeploy" layer { name: "data" type: "Input" top: "data" input_param { shape: { dim: 64 dim: 1 dim: 28 dim: 28 } } } layer { name: "conv2d_1" type: "Convolution" bottom: "data" top: "conv2d_1" convolution_param { num_output: 32 kernel_size: 3 stride: 1 } } layer { name: "relu_1" type: "ReLU" bottom: "conv2d_1" top: "conv2d_1" # inplace computation } layer { name: "max_pooling2d_1" type: "Pooling" bottom: "conv2d_1" top: "max_pooling2d_1" pooling_param { pool: MAX kernel_size: 3 stride: 2 } } layer { name: "conv2d_2" type: "Convolution" bottom: "max_pooling2d_1" top: "conv2d_2" convolution_param { num_output: 32 kernel_size: 3 stride: 1 } } layer { name: "relu_2" type: "ReLU" bottom: "conv2d_2" top: "conv2d_2" # inplace computation } layer { name: "dropout_1" type: "Dropout" bottom: "conv2d_2" top: "dropout_1" } layer { name: "dense_1" type: "InnerProduct" inner_product_param { num_output: 64 } bottom: "dropout_1" top: "dense_1" } layer { name: "relu_3" type: "ReLU" bottom: "dense_1" top: "dense_1" # inplace computation } layer { name: "dropout_2" type: "Dropout" bottom: "dense_1" top: "dropout_2" } layer { name: "dense_2" type: "InnerProduct" inner_product_param { num_output: 10 } bottom: "dropout_2" top: "dense_2" } layer { name: "prob" type: "Softmax" bottom: "dense_2" top: "prob" }

Load the trained model and check that the inference is correct.



In [6]:

    
model_def = 'example_caffe_mnist_model_deploy.prototxt'
model_weights = 'mnist.caffemodel'
deploy_net = caffe.Net(model_def,
                       model_weights,
                       caffe.TEST)
deploy_net.blobs









    Out[6]:





OrderedDict([('data', <caffe._caffe.Blob at 0x7fcfbf777ea0>),
             ('conv2d_1', <caffe._caffe.Blob at 0x7fcfbf777608>),
             ('max_pooling2d_1', <caffe._caffe.Blob at 0x7fcfbf777500>),
             ('conv2d_2', <caffe._caffe.Blob at 0x7fcfbf777660>),
             ('dropout_1', <caffe._caffe.Blob at 0x7fcfbf74b6b8>),
             ('dense_1', <caffe._caffe.Blob at 0x7fcfbf74b7c0>),
             ('dropout_2', <caffe._caffe.Blob at 0x7fcfbf74b710>),
             ('dense_2', <caffe._caffe.Blob at 0x7fcfbf7b4768>),
             ('prob', <caffe._caffe.Blob at 0x7fcfbf7b46b8>)])



In [14]:

    
deploy_net.blobs['data'].data[...] = seven
output = deploy_net.forward()
output['prob'][0].argmax()









    Out[14]:





7



In [ ]:



In [ ]: