| # -*- coding: utf-8 -*- """ Created on Mon Apr 20 16:55:55 2015 @author: 陳日偉 <riwei.chen@outlook.com> @brief：在lfw數(shù)據(jù)庫(kù)上驗(yàn)證訓(xùn)練好了的網(wǎng)絡(luò) """ import sklearn import numpy as np import matplotlib.pyplot as plt import skimage caffe_root = '/home/crw/caffe-master/' import sys sys.path.insert(0, caffe_root + 'python') import caffe import sklearn.metrics.pairwise as pw def read_imagelist(filelist): ''' @brief：從列表文件中，讀取圖像數(shù)據(jù)到矩陣文件中 @param： filelist 圖像列表文件 @return ：4D 的矩陣 ''' fid=open(filelist) lines=fid.readlines() test_num=len(lines) fid.close() X=np.empty((test_num,3,64,64)) i =0 for line in lines: word=line.split('\n') filename=word[0] im1=skimage.io.imread(filename,as_grey=False) image =skimage.transform.resize(im1,(64, 64))*255 if image.ndim<3: print 'gray:'+filename X[i,0,:,:]=image[:,:] X[i,1,:,:]=image[:,:] X[i,2,:,:]=image[:,:] else: X[i,0,:,:]=image[:,:,0] X[i,1,:,:]=image[:,:,1] X[i,2,:,:]=image[:,:,2] i=i+1 return X def read_labels(labelfile): ''' 讀取標(biāo)簽列表文件 ''' fin=open(labelfile) lines=fin.readlines() labels=np.empty((len(lines),)) k=0; for line in lines: labels[k]=int(line) k=k+1; fin.close() return labels def draw_roc_curve(fpr,tpr,title='cosine',save_name='roc_lfw'): ''' 畫(huà)ROC曲線圖 ''' plt.figure() plt.plot(fpr, tpr) plt.plot([0, 1], [0, 1], 'k--') plt.xlim([0.0, 1.0]) plt.ylim([0.0, 1.0]) plt.xlabel('False Positive Rate') plt.ylabel('True Positive Rate') plt.title('Receiver operating characteristic using: '+title) plt.legend(loc="lower right") plt.show() plt.savefig(save_name+'.png') def evaluate(itera=500000,metric='cosine'): ''' @brief: 評(píng)測(cè)模型的性能 @param：itera： 模型的迭代次數(shù) @param：metric： 度量的方法 ''' # 轉(zhuǎn)換均值圖像數(shù)據(jù) -->npy格式文件 fin='/media/crw/MyBook/TrainData/LMDB/CASIA-WebFace/10575_64X64/mean.binaryproto' fout='/media/crw/MyBook/TrainData/LMDB/CASIA-WebFace/10575_64X64/mean.npy' blob = caffe.proto.caffe_pb2.BlobProto() data = open( fin , 'rb' ).read() blob.ParseFromString(data) arr = np.array( caffe.io.blobproto_to_array(blob) ) out = arr[0] np.save( fout , out ) #設(shè)置為gpu格式 caffe.set_mode_gpu() net = caffe.Classifier('/home/crw/caffe-master/FaceRecognition/try5_2/deploy.prototxt', '/media/crw/MyBook/Model/FaceRecognition/try5_2/snapshot_iter_'+str(itera)+'.caffemodel', mean=np.load(fout)) #需要對(duì)比的圖像，一一對(duì)應(yīng) filelist_left='./LFW_Test_List/left.list' filelist_right='./LFW_Test_List/right.list' filelist_label='./LFW_Test_List/label.list' print 'network input :' ,net.inputs print 'network output： ', net.outputs #提取左半部分的特征 X=read_imagelist(filelist_left) test_num=np.shape(X)[0] #data_1 是輸入層的名字 out = net.forward_all(data_1 = X) feature1 = np.float64(out['deepid_1']) feature1=np.reshape(feature1,(test_num,160)) #np.savetxt('feature1.txt', feature1, delimiter=',') #提取右半部分的特征 X=read_imagelist(filelist_right) out = net.forward_all(data_1=X) feature2 = np.float64(out['deepid_1']) feature2=np.reshape(feature2,(test_num,160)) #np.savetxt('feature2.txt', feature2, delimiter=',') #提取標(biāo)簽 labels=read_labels(filelist_label) assert(len(labels)==test_num) #計(jì)算每個(gè)特征之間的距離 mt=pw.pairwise_distances(feature1, feature2, metric=metric) predicts=np.empty((test_num,)) for i in range(test_num): predicts[i]=mt[i][i] # 距離需要?dú)w一化到0--1,與標(biāo)簽0-1匹配 for i in range(test_num): predicts[i]=(predicts[i]-np.min(predicts))/(np.max(predicts)-np.min(predicts)) print 'accuracy is :',calculate_accuracy(predicts,labels,test_num) np.savetxt('predict.txt',predicts) fpr, tpr, thresholds=sklearn.metrics.roc_curve(labels,predicts) draw_roc_curve(fpr,tpr,title=metric,save_name='lfw_'+str(itera)) def calculate_accuracy(distance,labels,num): ''' #計(jì)算識(shí)別率, 選取閾值，計(jì)算識(shí)別率 ''' accuracy = [] predict = np.empty((num,)) threshold = 0.2 while threshold <= 0.8 : for i in range(num): if distance[i] >= threshold: predict[i] =1 else: predict[i] =0 predict_right =0.0 for i in range(num): if predict[i]==labels[i]: predict_right = 1.0+predict_right current_accuracy = (predict_right/num) accuracy.append(current_accuracy) threshold=threshold+0.001 return np.max(accuracy) if __name__=='__main__': itera=500000 metric='cosine' evaluate(itera,metric) |
|
|
