caffe（9）使用jupyter測試單張圖片代碼實例

首先，在命令行打開jupyter

sudo ipython notebook --allow-root

Then, open a new python, you can see like this

matplotlib---paint the pictures

#import required packagesimport numpy as npimport matplotlib.pyplot as pltimport os,sys,caffefrom PIL import Image%matplotlib inline#set path to rootcaffe_root='/home/peipei/caffe/'os.chdir(caffe_root)sys.path.insert(0,caffe_root+'python')#set show parametersplt.rcParams['figure.figsize'] = (64, 64)plt.rcParams['image.interpolation'] = 'nearest'plt.rcParams['image.cmap'] = 'gray'#import caffe model, set the model architecture deploy.prototxt and model parameters .caffemodelnet = caffe.Net(caffe_root + 'examples/myfile2/cifar10_quick.prototxt',                caffe_root + 'examples/myfile2/cifar10_quick_iter_300.caffemodel',                caffe.TEST)#print each layer[(k, v[0].data.shape) for k, v in net.params.items()]net.blobs['data'].data.shape#编写一个函数，用于显示各层的参数def show_feature(data, padsize=1, padval=0):    data -= data.min()    data /= data.max()        # 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, padsize), (0, padsize)) + ((0, 0),) * (data.ndim - 3)    data = np.pad(data, padding, mode='constant', constant_values=(padval, padval))        # 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, cmap='gray')    plt.axis('off')# first convolution layer, show in imagesweight = net.params["conv1"][0].dataprint weight.shape#show_feature(weight.transpose(0, 2, 3, 1))#set CPU modecaffe.set_mode_cpu()#加载测试图片，并显示im = caffe.io.load_image('/home/peipei/caffe/examples/myfile2/test/FVC2004_fingerprints_544/4/ms_ds2_s0001_fo10.bmp', False)print im.shape#imr = im.resize((64,64)) #resize image required by inputprint im.shape#plt.imshow(im)#　编写一个函数，将二进制的均值转换为python的均值def convert_mean(binMean,npyMean):    blob = caffe.proto.caffe_pb2.BlobProto()    bin_mean = open(binMean, 'rb' ).read()    blob.ParseFromString(bin_mean)    arr = np.array( caffe.io.blobproto_to_array(blob) )    npy_mean = arr[0]    np.save(npyMean, npy_mean )#generate the python file: mean.npybinMean=caffe_root+'examples/myfile2/mean.binaryproto'npyMean=caffe_root+'examples/myfile2/mean.npy'convert_mean(binMean,npyMean)#将图片载入blob中,并减去均值transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape}) # 设定图片的shape格式(1,1,64,64)  transformer.set_transpose('data', (2,0,1)) # 改变维度的顺序，由原始图片(64,64,1)变为(1,64,64)transformer.set_mean('data', np.load(npyMean).mean(1).mean(1)) # 减去均值transformer.set_raw_scale('data', 255)  # 缩放到【0，255】之间  #transformer.set_channel_swap('data', (2,1,0))  #交换通道，将图片由RGB变为BGR  net.blobs['data'].data[...] = transformer.preprocess('data',im) #执行上面设置的图片预处理操作，并将图片载入到blob中 inputData=net.blobs['data'].dataout = net.forward()#　编写一个函数，用于显示各层数据def show_data(data, padsize=1, padval=0):    data -= data.min()    data /= data.max()        # 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, padsize), (0, padsize)) + ((0, 0),) * (data.ndim - 3)    data = np.pad(data, padding, mode='constant', constant_values=(padval, padval))        # 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.figure()    plt.imshow(data,cmap='gray')    plt.axis('off')plt.rcParams['figure.figsize'] = (64, 64)plt.rcParams['image.interpolation'] = 'nearest'plt.rcParams['image.cmap'] = 'gray'#显示第一个卷积层的输出数据和权值（filter）show_data(net.blobs['conv1'].data[0])print net.blobs['conv1'].data.shapeshow_data(net.params['conv1'][0].data.reshape(32,5,5))print net.params['conv1'][0].data.shape# 最后一层输入属于某个类的概率prob = net.blobs['prob'].data[0].flatten()print probplt.plot(prob) order=prob.argsort()[2]print order

finanly, show the probability of all the labels:

[ 0.02902012  0.946181    0.02479883]

print order show the label with the highest probability.