images[0:1000, :, :, :], labels[0:1000], test_size=0.1)
class_number = len(names) + 1
images_train = images_train.astype('float32')
images_test = images_test.astype('float32')
images_train = images_train / 255.0
images_test = images_test / 255.0
# the labels are converted to binary class matrices
labels_test_b = keras.utils.to_categorical(labels_test, class_number)
labels_train_b = keras.utils.to_categorical(labels_train, class_number)
'''-----------training and testing--------------'''
'''RMSprop'''
#train
start_time = time.time()
optimizer = keras.optimizers.RMSprop(learning_rate=0.0001, decay=1e-6)
model = CNN.CNNmodel(batchsize=32,
epochs=100,
classnumber=class_number,
optimizer=optimizer)
model.build_model()
model.get_traindata(images_train, labels_train_b)
model.train_model()
train_time = time.time() - start_time
print("RMSprop train_time:%s seconds" % (train_time))
model.save_model('RMSprop3')
# save the loss and accuracy of training
model.save_para('TrainAcc_RMS3', 'TrainLoss_RMS3')
#test
start_time = time.time()
model.get_testdata(images_test, labels_test_b)
test_loss, test_acc = model.evaluate()
test_time = time.time() - start_time
