KNN_KNN分类_CIFAR-10_knn图像分类

import cPickle from KNN_1 import * import numpy as np def load_cifar10(filename): with open(filename, 'rb') as fo: data = cPickle.load(fo) return data cifar_train = {} for i in range(5): cifar1 = load_cifar10('data_batch_'+str(i+1)) if i == 0: cifar_train[b'data'] = cifar1[b'data'] cifar_train[b'labels'] = cifar1[b'labels'] else: cifar_train[b'data'] = np.vstack([cifar_train[b'data'], cifar1[b'data']]) cifar_train[b'labels'] = np.hstack([cifar_train[b'labels'], cifar1[b'labels']]) target_name = load_cifar10('batches.meta') cifar_train[b'label_names'] = target_name[b'label_names'] cifar_test = load_cifar10('test_batch') cifar_test[b'labels'] = np.array(cifar_test[b'labels']) x_train = cifar_train[b'data'] y_train = cifar_train[b'labels'] x_test = cifar_test[b'data'] y_test = cifar_test[b'labels'] x_train_rows = x_train.reshape(x_train.shape[0], 32*32*3) x_test_rows = x_test.reshape(x_test.shape[0], 32*32*3) x_train_rows_fake = x_train_rows[5000:, :] y_train_fake = y_train[5000:] x_test_rows_fake = x_train_rows[:5000, :] y_test_fake = y_train[:5000] accuracy_k = np.zeros(5, dtype=float) i1 = 0 kk = [1, 3, 5, 7, 9] print(kk) for k in kk: nn = NearestNeighbor() nn.train(x_train_rows_fake, y_train_fake) Yte_predict = nn.predict(x_test_rows_fake, k=k) accuracy_k[i1] = np.mean(Yte_predict == y_test_fake) print(accuracy_k[i1]) i1 = i1 + 1 print('accuracy_k:') print(accuracy_k) k = kk[np.argmax(accuracy_k)] print'k=%d' % k nn = NearestNeighbor() nn.train(x_train_rows, y_train) Yte_predict = nn.predict(x_test_rows, k=k) print'accuracy:%f' % (np.mean(Yte_predict == y_test))