Shulou(Shulou.com)06/03 Report--

Dataset introduction

LFW (Labeled Faces in the Wild) face database is a database compiled by the computer Vision Laboratory of the University of Massachusetts Amherst, which is mainly used to study the problem of face recognition under unrestricted conditions. The LFW database, which mainly collects images from the Internet, rather than laboratories, contains more than 13000 face images, each of which is identified by the corresponding person's name, of which 1680 people correspond to more than one image, that is, about 1680 people contain more than two faces. LFW dataset mainly tests the accuracy of face recognition.

The code implements from time import time # record time import logging # run log of print program import matplotlib.pyplot as pltfrom sklearn.cross_validation import train_test_split # partition training set and test set from sklearn.datasets import fetch_lfw_people # import data set (celebrity) from sklearn.grid_search import GridSearchCV # parameter from sklearn of debug function .metrics import classification_report # displays classified reports Show the main classification indicators, accuracy, recall rate and F1 score from sklearn.metrics import confusion_matrix # to judge the real and predicted categories The svc equation from sklearn.cluster.tests.test_k_means import n_samples# training set of from sklearn.decomposition import RandomizedPCA # pca reduced-dimensional from sklearn.svm import SVC # svm is expressed in matrix form: sklearn's own face image data set lfw_people = fetch_lfw_people (min_faces_per_person=70,resize=0.4) nasty samplesh W = number of lfw_people.images.shape # instances, h, wx = lfw_people.data # all training data, 1288 pictures 1850 eigenvalues per picture n_features = x.shape [1] # Dimension of feature vector 1850y = lfw_people.target # face tag target_names = lfw_people.target_names # Human name n_class = target_names.shape [0] # several people need to recognize print ("total dataset size" : ") print (" n_samples:% d "% n_samples) print (" n_features:% d "% n_features) print (" n_class:% d "% n_class) x_train The high-dimensional eigenvector is randomly reduced to the low-dimensional one. First, the model pca = RandomizedPCA is established. Fit (x_train) print ("done in% 0.3fs"% (time ()-t0)) # extract the eigenvalues of the human face eigenfaces = pca.components_.reshape ((nasty 0.3fs)% (time ()-t0) W)) # transform facial features into feature vector set print ('projecting the input data on the eigenfaces orthonomal basis') t0 = time () # to reduce the dimension of the data model Reduced to 150x_train_pca = pca.transform (x_train) x_test_pca = pca.transform (x_test) print ("done in% 0.3fs"% (time ()-t0)) print ("fitting the classfier to the training set") t0 = time () # C is the penalty for the wrong part Gamma compositing point para_grid = {'gamma: [1e3Power5e3Power1e4magic5e4Power1e5],' gamma':],} # rbf deals with images better, C and gamma combine, name the best combination using GridSearchCV for free combination. Finally determine the appropriate combination clf = GridSearchCV (SVC (kernel='rbf'), para_grid) clf = clf.fit (x_train_pca) Y_train) print ("done in% 0.3fs"% (time ()-t0)) print ("best estimator found by grid search:") print (clf.best_estimator_) # Information about the best model print ("predict the people's name on the test set") t0 = time () y_pred = clf.predict (x_test_pca) print ("done in% 0.3fs" % (time ()-t0) print (classification_report (y_test) ) print (confusion_matrix) def plot_gallery (images,titles,h,w,n_row = 3) images,titles,h,w,n_row = 4): plt.figure (figsize= (1.8 minutes) 2.4 minutes row)) plt.subplots_adjust (bottom = 0. 01 camera right =. 99. 90) Hspace = .35) for i in range (n_row * n_col): plt.subplot plt.imshow (images [I] .reshape ((hjournal w)), cmap=plt.cm.gray) plt.title (titles [I], size = 12) plt.xticks () plt.yticks () def title I): pred_name = target_ namespace [y _ pred [I]] .rsplit (', 1) [- 1] true_name = target_ [y _ test [I] .rsplit (', 1) [- 1] return 'predicted:% s\ nture:% s'% (pred_name,true_name) prediction_titles = [title I) for i in range (y_pred.shape [0])] plot_gallery (xonometer) eigenface_title = ["eigenface% d"% i for i in range (eigenfaces.shape [0])] plot_gallery (eigenfaces,eigenface_title,h W) plt.show () result total dataset size:n_samples: 1288n_features: 1850n_class: 7done in 0.270sprojecting the input data on the eigenfaces orthonomal basisdone in 0.040sfitting the classfier to the training setdone in 30.796sbest estimator found by grid search:SVC (cache_size=200 1000.0, cache_size=200, class_weight=None, coef0=0.0, decision_function_shape='ovr', degree=3, gamma=0.005, kernel='rbf', max_iter=-1, probability=False, random_state=None, shrinking=True, tol=0.001 Verbose=False) predict the people's name on the test setdone in 0.125s precision recall f1-score supportAriel Sharon 0.91 0.67 0.77 15Colin Powell 0.80 0.82 0.81 45Donald Rumsfeld 0.96 0.61 0.75 41George W Bush 0 . 79 0.96 0.87 144Gerhard Schroeder 0.95 0.63 0.76 30Hugo Chavez 1.00 0.79 0.88 19Tony Blair 0.86 0.89 0.88 28avg / total 0.85 0.84 0.83 322 Confusion matrix [[10 20 3 000] [1 37 0 7 000] [0 1 25 13 10 1] [0 4 1 138 0 0 0 1] [0 10 8 19 0 2] [0 103 0 15 0] [0 000 3 0 25]]

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.