Keras : Vision models サンプル: mnist_cnn.py
MNIST データセット上で単純な ConvNet をトレーニングします。
12 エポック後に 99.25 % テスト精度を得ます。50 エポック後では 99.21 % テスト精度でした。
from __future__ import print_function import keras from keras.datasets import mnist from keras.models import Sequential from keras.layers import Dense, Dropout, Flatten from keras.layers import Conv2D, MaxPooling2D from keras import backend as K
batch_size = 128 num_classes = 10 epochs = 50
# input image dimensions img_rows, img_cols = 28, 28
# the data, shuffled and split between train and test sets (x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1) x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1) input_shape = (img_rows, img_cols, 1)
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255
print('x_train shape:', x_train.shape)
print(x_train.shape[0], 'train samples')
print(x_test.shape[0], 'test samples')
x_train shape: (60000, 28, 28, 1) 60000 train samples 10000 test samples
# convert class vectors to binary class matrices y_train = keras.utils.to_categorical(y_train, num_classes) y_test = keras.utils.to_categorical(y_test, num_classes)
model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3),
activation='relu',
input_shape=input_shape))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(num_classes, activation='softmax'))
model.summary()
_________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_1 (Conv2D) (None, 26, 26, 32) 320 _________________________________________________________________ conv2d_2 (Conv2D) (None, 24, 24, 64) 18496 _________________________________________________________________ max_pooling2d_1 (MaxPooling2 (None, 12, 12, 64) 0 _________________________________________________________________ dropout_1 (Dropout) (None, 12, 12, 64) 0 _________________________________________________________________ flatten_1 (Flatten) (None, 9216) 0 _________________________________________________________________ dense_1 (Dense) (None, 128) 1179776 _________________________________________________________________ dropout_2 (Dropout) (None, 128) 0 _________________________________________________________________ dense_2 (Dense) (None, 10) 1290 ================================================================= Total params: 1,199,882 Trainable params: 1,199,882 Non-trainable params: 0
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.Adadelta(),
metrics=['accuracy'])
model.fit(x_train, y_train,
batch_size=batch_size,
epochs=epochs,
verbose=1,
validation_data=(x_test, y_test))
Train on 60000 samples, validate on 10000 samples Epoch 1/50 60000/60000 [==============================] - 14s 240us/step - loss: 0.3250 - acc: 0.9003 - val_loss: 0.0804 - val_acc: 0.9745 Epoch 2/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.1135 - acc: 0.9660 - val_loss: 0.0538 - val_acc: 0.9823 Epoch 3/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0851 - acc: 0.9739 - val_loss: 0.0432 - val_acc: 0.9858 Epoch 4/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0721 - acc: 0.9785 - val_loss: 0.0400 - val_acc: 0.9865 Epoch 5/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0627 - acc: 0.9812 - val_loss: 0.0367 - val_acc: 0.9874 Epoch 6/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0552 - acc: 0.9834 - val_loss: 0.0331 - val_acc: 0.9892 Epoch 7/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0519 - acc: 0.9850 - val_loss: 0.0315 - val_acc: 0.9889 Epoch 8/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0472 - acc: 0.9860 - val_loss: 0.0287 - val_acc: 0.9904 Epoch 9/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0454 - acc: 0.9867 - val_loss: 0.0299 - val_acc: 0.9899 Epoch 10/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0410 - acc: 0.9877 - val_loss: 0.0294 - val_acc: 0.9895 Epoch 11/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0383 - acc: 0.9888 - val_loss: 0.0286 - val_acc: 0.9901 Epoch 12/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0378 - acc: 0.9880 - val_loss: 0.0284 - val_acc: 0.9907 Epoch 13/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0361 - acc: 0.9892 - val_loss: 0.0278 - val_acc: 0.9912 Epoch 14/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0338 - acc: 0.9903 - val_loss: 0.0267 - val_acc: 0.9915 Epoch 15/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0333 - acc: 0.9899 - val_loss: 0.0274 - val_acc: 0.9911 Epoch 16/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0331 - acc: 0.9899 - val_loss: 0.0304 - val_acc: 0.9903 Epoch 17/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0309 - acc: 0.9904 - val_loss: 0.0263 - val_acc: 0.9920 Epoch 18/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0306 - acc: 0.9911 - val_loss: 0.0321 - val_acc: 0.9905 Epoch 19/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0309 - acc: 0.9909 - val_loss: 0.0295 - val_acc: 0.9908 Epoch 20/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0291 - acc: 0.9914 - val_loss: 0.0303 - val_acc: 0.9901 Epoch 21/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0288 - acc: 0.9913 - val_loss: 0.0280 - val_acc: 0.9909 Epoch 22/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0292 - acc: 0.9913 - val_loss: 0.0291 - val_acc: 0.9915 Epoch 23/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0277 - acc: 0.9920 - val_loss: 0.0304 - val_acc: 0.9906 Epoch 24/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0289 - acc: 0.9915 - val_loss: 0.0268 - val_acc: 0.9919 Epoch 25/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0290 - acc: 0.9915 - val_loss: 0.0282 - val_acc: 0.9910 Epoch 26/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0278 - acc: 0.9914 - val_loss: 0.0272 - val_acc: 0.9920 Epoch 27/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0270 - acc: 0.9921 - val_loss: 0.0282 - val_acc: 0.9915 Epoch 28/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0266 - acc: 0.9923 - val_loss: 0.0290 - val_acc: 0.9919 Epoch 29/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0267 - acc: 0.9919 - val_loss: 0.0321 - val_acc: 0.9894 Epoch 30/50 60000/60000 [==============================] - 10s 174us/step - loss: 0.0278 - acc: 0.9919 - val_loss: 0.0259 - val_acc: 0.9921 Epoch 31/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0270 - acc: 0.9917 - val_loss: 0.0306 - val_acc: 0.9912 Epoch 32/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0271 - acc: 0.9922 - val_loss: 0.0312 - val_acc: 0.9908 Epoch 33/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0262 - acc: 0.9922 - val_loss: 0.0306 - val_acc: 0.9918 Epoch 34/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0266 - acc: 0.9921 - val_loss: 0.0330 - val_acc: 0.9908 Epoch 35/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0269 - acc: 0.9917 - val_loss: 0.0295 - val_acc: 0.9919 Epoch 36/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0264 - acc: 0.9920 - val_loss: 0.0286 - val_acc: 0.9913 Epoch 37/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0285 - acc: 0.9916 - val_loss: 0.0302 - val_acc: 0.9909 Epoch 38/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0266 - acc: 0.9921 - val_loss: 0.0307 - val_acc: 0.9914 Epoch 39/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0270 - acc: 0.9922 - val_loss: 0.0293 - val_acc: 0.9919 Epoch 40/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0262 - acc: 0.9921 - val_loss: 0.0313 - val_acc: 0.9917 Epoch 41/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0259 - acc: 0.9922 - val_loss: 0.0307 - val_acc: 0.9909 Epoch 42/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0271 - acc: 0.9920 - val_loss: 0.0330 - val_acc: 0.9910 Epoch 43/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0266 - acc: 0.9923 - val_loss: 0.0327 - val_acc: 0.9913 Epoch 44/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0285 - acc: 0.9917 - val_loss: 0.0299 - val_acc: 0.9917 Epoch 45/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0254 - acc: 0.9928 - val_loss: 0.0301 - val_acc: 0.9910 Epoch 46/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0254 - acc: 0.9923 - val_loss: 0.0292 - val_acc: 0.9914 Epoch 47/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0273 - acc: 0.9923 - val_loss: 0.0327 - val_acc: 0.9912 Epoch 48/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0276 - acc: 0.9919 - val_loss: 0.0319 - val_acc: 0.9914 Epoch 49/50 60000/60000 [==============================] - 10s 172us/step - loss: 0.0255 - acc: 0.9925 - val_loss: 0.0325 - val_acc: 0.9911 Epoch 50/50 60000/60000 [==============================] - 10s 173us/step - loss: 0.0255 - acc: 0.9924 - val_loss: 0.0296 - val_acc: 0.9921 CPU times: user 8min 3s, sys: 50.4 s, total: 8min 53s Wall time: 8min 43s
score = model.evaluate(x_test, y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])
Test loss: 0.0296470036264 Test accuracy: 0.9921
以上