本文將會介紹Caffe的Python接口的使用方法。編輯Python可以使用很多種方法，我們採用的是IPython交互式編輯環境。
1 Python的安裝
如果你的Windows電腦還沒有安裝Python，請先自行搜索Python的安裝方法，例如 http://jupyter.org/install.html，推薦使用Anaconda軟件包安裝方式，這樣就自帶IPython/Jupyter環境了。本文使用的是Python2.7。
2 Caffe的安裝
Windows Caffe的安裝請參照之前的一篇文章：
http://blog.csdn.net/zzlyw/article/details/66971669
3 詳細操作
3.1 設置
（1）首先，設置Python、numpy、和matplotlib。
In [1]:
# set up Python environment: numpy for numerical routines, and matplotlib for plotting
import numpy as np
import matplotlib.pyplot as plt
# display plots in this notebook
get_ipython().magic(u'matplotlib inline')
# set display defaults
plt.rcParams['figure.figsize'] = (10, 10) # large images
plt.rcParams['image.interpolation'] = 'nearest' # don't interpolate: show square pixels
plt.rcParams['image.cmap'] = 'gray' # use grayscale output rather than a (potentially misleading) color heatmap
（2）導入caffe
In [2]:
# The caffe module needs to be on the Python path;
# we'll add it here explicitly.
import sys
caffe_root = 'F:\\Projects\\caffe\\' # this file should be run from {caffe_root}/examples (otherwise change this line)
sys.path.insert(0, caffe_root + 'python')
import caffe
# If you get "No module named _caffe", either you have not built pycaffe or you have the wrong path.（3）如果還沒有自己訓練好的模型，可以下載一個CaffeNet
In [3]:
import os
if os.path.isfile(caffe_root + 'models\\bvlc_reference_caffenet\\bvlc_reference_caffenet.caffemodel'):
print 'CaffeNet found.'
else:
print 'Downloading pre-trained CaffeNet model...'
get_ipython().system(u'python F:\\Projects\\caffe\\scripts\\download_model_binary.py F:\\Projects\\caffe\\models\\bvlc_reference_caffenet')
Out:
CaffeNet found.
3.2 導入網絡和輸入預處理
（1）設置Caffe爲CPU模式，從硬盤導入網絡。
In [4]:caffe.set_mode_cpu()
model_def = caffe_root + 'models\\bvlc_reference_caffenet\\deploy.prototxt'
model_weights = caffe_root + 'models\\bvlc_reference_caffenet\\bvlc_reference_caffenet.caffemodel'
net = caffe.Net(model_def, # defines the structure of the model
model_weights, # contains the trained weights
caffe.TEST) # use test mode (e.g., don't perform dropout)
（2）設置輸入預處理。我們使用Caffe的caffe.io.Transformer 來做這件事，它與caffe的其他部分是獨立的，所以任何其他自定義的預處理代碼都可以使用。
默認的CaffeNet使用圖像爲BGR格式。它們的灰度範圍應該使用[0 , 255]，於是可以使用ImageNet的圖像像素均值作爲要減去的數值。
Matplotlib會把導入的圖像設定爲[0, 1]範圍的RGB格式，所以需要做一些轉換。
In [5]:
# load the mean ImageNet image (as distributed with Caffe) for subtraction
mu = np.load(caffe_root + 'python/caffe/imagenet/ilsvrc_2012_mean.npy')
mu = mu.mean(1).mean(1) # average over pixels to obtain the mean (BGR) pixel values
print 'mean-subtracted values:', zip('BGR', mu)
# create transformer for the input called 'data'
transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})
transformer.set_transpose('data', (2,0,1)) # move image channels to outermost dimension
transformer.set_mean('data', mu) # subtract the dataset-mean value in each channel
transformer.set_raw_scale('data', 255) # rescale from [0, 1] to [0, 255]
transformer.set_channel_swap('data', (2,1,0)) # swap channels from RGB to BGR
Out:
3.3 CPU分類
（1）設置batch size爲50
In [6]:
# set the size of the input (we can skip this if we're happy
# with the default; we can also change it later, e.g., for different batch sizes)
net.blobs['data'].reshape(50, # batch size
3, # 3-channel (BGR) images
227, 227) # image size is 227x227（2）導入圖像，執行預處理
In [7]:image = caffe.io.load_image(caffe_root + 'examples/images/cat.jpg')
transformed_image = transformer.preprocess('data', image)
plt.imshow(image)Out:
<matplotlib.image.AxesImage at 0x19ed6ba8>
（3）執行分類
In [8]:# copy the image data into the memory allocated for the net
net.blobs['data'].data[...] = transformed_image
### perform classification
output = net.forward()
output_prob = output['prob'][0] # the output probability vector for the first image in the batch
print 'predicted class is:', output_prob.argmax()Out:
predicted class is: 281
（4）網絡給出了一個概率向量，最可能的類別是編號281的類。我們需要找到Image的類別標籤。下面的程序是檢驗有沒有sysset_words.txt文件，如果沒有則使用腳本從網上下載。由於腳本本來是在Linux shell中運行的，在Windows命令行中執行報錯，所以我是先使用別的方法下載了這個文件，放到了該對應的路徑下。你可以使用win10自帶的Linux內核系統運行shell命令來下載，也可以從網上搜索這個文件。
In [9]:# load ImageNet labels
labels_file = caffe_root + 'data\\ilsvrc12\\synset_words.txt'
if not os.path.exists(labels_file):
get_ipython().system(u'F:\Projects\caffe\data\ilsvrc12\get_ilsvrc_aux.sh')
labels = np.loadtxt(labels_file, str, delimiter='\t')
print 'output label:', labels[output_prob.argmax()]（5）查看全部分類結果列表
In [10]:# sort top five predictions from softmax output
top_inds = output_prob.argsort()[::-1][:5] # reverse sort and take five largest items
print 'probabilities and labels:'
zip(output_prob[top_inds], labels[top_inds])Out:
3.4 使用GPU模式
（1）先看下CPU模式下分類時間
In [11]:get_ipython().magic(u'timeit net.forward()')Out:
（2）改到GPU模式下看分類時間
In [12]:caffe.set_device(0) # if we have multiple GPUs, pick the first one
caffe.set_mode_gpu()
net.forward() # run once before timing to set up memory
get_ipython().magic(u'timeit net.forward()')10 loops, best of 3: 51.9 ms per loop
3.5 檢查中間輸出
網絡並非是一個黑盒，讓我們看看中間的參數信息。
In [13]:# for each layer, show the output shape
for layer_name, blob in net.blobs.iteritems():
print layer_name + '\t' + str(blob.data.shape)data (50L, 3L, 227L, 227L) conv1 (50L, 96L, 55L, 55L) pool1 (50L, 96L, 27L, 27L) norm1 (50L, 96L, 27L, 27L) conv2 (50L, 256L, 27L, 27L) pool2 (50L, 256L, 13L, 13L) norm2 (50L, 256L, 13L, 13L) conv3 (50L, 384L, 13L, 13L) conv4 (50L, 384L, 13L, 13L) conv5 (50L, 256L, 13L, 13L) pool5 (50L, 256L, 6L, 6L) fc6 (50L, 4096L) fc7 (50L, 4096L) fc8 (50L, 1000L) prob (50L, 1000L)
In [14]:
for layer_name, param in net.params.iteritems():
print layer_name + '\t' + str(param[0].data.shape), str(param[1].data.shape)conv1 (96L, 3L, 11L, 11L) (96L,) conv2 (256L, 48L, 5L, 5L) (256L,) conv3 (384L, 256L, 3L, 3L) (384L,) conv4 (384L, 192L, 3L, 3L) (384L,) conv5 (256L, 192L, 3L, 3L) (256L,) fc6 (4096L, 9216L) (4096L,) fc7 (4096L, 4096L) (4096L,) fc8 (1000L, 4096L) (1000L,)
def vis_square(data):
"""Take an array of shape (n, height, width) or (n, height, width, 3)
and visualize each (height, width) thing in a grid of size approx. sqrt(n) by sqrt(n)"""
# normalize data for display
data = (data - data.min()) / (data.max() - data.min())
# force the number of filters to be square
n = int(np.ceil(np.sqrt(data.shape[0])))
padding = (((0, n ** 2 - data.shape[0]),
(0, 1), (0, 1)) # add some space between filters
+ ((0, 0),) * (data.ndim - 3)) # don't pad the last dimension (if there is one)
data = np.pad(data, padding, mode='constant', constant_values=1) # pad with ones (white)
# tile the filters into an image
data = data.reshape((n, n) + data.shape[1:]).transpose((0, 2, 1, 3) + tuple(range(4, data.ndim + 1)))
data = data.reshape((n * data.shape[1], n * data.shape[3]) + data.shape[4:])
plt.imshow(data); plt.axis('off')In [16]:
# the parameters are a list of [weights, biases]
filters = net.params['conv1'][0].data
vis_square(filters.transpose(0, 2, 3, 1))Out:
In [17]:
feat = net.blobs['conv1'].data[0, :36]
vis_square(feat)
In [18]:
feat = net.blobs['pool5'].data[0]
vis_square(feat)
In [19]:
feat = net.blobs['fc6'].data[0]
plt.subplot(2, 1, 1)
plt.plot(feat.flat)
plt.subplot(2, 1, 2)
_ = plt.hist(feat.flat[feat.flat > 0], bins=100)
In [20]:
feat = net.blobs['prob'].data[0]
plt.figure(figsize=(15, 3))
plt.plot(feat.flat)
3.6 嘗試自己的圖像
In [21]:# download an image
#my_image_url = "https://timgsa.baidu.com/timg?image&quality=80&size=b9999_10000&sec=1491715902209&di=82ef5c02c812e21e2e0f44fce2a1d4b6&imgtype=0&src=http%3A%2F%2Fcyjctrip.qiniudn.com%2F56329%2F1374595566800p18064d9kk169p1j291j1l1u31k0lk.jpg" # paste your URL here
# for example:
# my_image_url = "https://upload.wikimedia.org/wikipedia/commons/b/be/Orang_Utan%2C_Semenggok_Forest_Reserve%2C_Sarawak%2C_Borneo%2C_Malaysia.JPG"
#!wget -O image.jpg $my_image_url
# transform it and copy it into the net
image = caffe.io.load_image('C:\\Users\\Bill\\Desktop\\image.jpg')
net.blobs['data'].data[...] = transformer.preprocess('data', image)
# perform classification
net.forward()
# obtain the output probabilities
output_prob = net.blobs['prob'].data[0]
# sort top five predictions from softmax output
top_inds = output_prob.argsort()[::-1][:5]
plt.imshow(image)
print 'probabilities and labels:'
zip(output_prob[top_inds], labels[top_inds])Out:
[(0.69523662, 'n02403003 ox'), (0.16318876, 'n02389026 sorrel'), (0.039488554, 'n02087394 Rhodesian ridgeback'), (0.029075578, 'n03967562 plow, plough'), (0.015077997, 'n02422106 hartebeest')]
