2. Preprocessing
Google Colab에서 작업
Load Packages
from PIL import Image
import re
import pandas as pd
import numpy as np
from tqdm import tqdm
from keras import layers, models, optimizers, losses, metrics
from keras.callbacks import ModelCheckpoint
from keras.applications import resnet50, inception_v3
import matplotlib.pyplot as plt
tqdm().pandas()
0it [00:00, ?it/s]
Load Datasets
total_df = pd.read_csv('/content/drive/Shared drives/personal/dataframe/total_df.csv')
poster_total = np.load('/content/drive/Shared drives/personal/poster/poster_total.npy')
total_df
| code | title_kor | title_eng | year | rating | rank | link | genre | |
|---|---|---|---|---|---|---|---|---|
| 0 | 171539 | 그린 북 | Green Book | 2018 | 9.59 | 12세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['드라마'] |
| 1 | 174830 | 가버나움 | Capharnaum | 2018 | 9.58 | 15세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['드라마'] |
| 2 | 151196 | 원더 | Wonder | 2017 | 9.49 | 전체 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['드라마'] |
| 3 | 169240 | 아일라 | Ayla: The Daughter of War | 2017 | 9.48 | 15세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['드라마', '전쟁'] |
| 4 | 157243 | 당갈 | Dangal | 2016 | 9.47 | 12세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['드라마', '액션'] |
| ... | ... | ... | ... | ... | ... | ... | ... | ... |
| 1395 | 41334 | 옹박 - 두번째 미션 | The Protector | 2005 | 8.28 | 15세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['액션', '범죄', '드라마', '스릴러'] |
| 1396 | 162249 | 램페이지 | RAMPAGE | 2018 | 8.28 | 12세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['액션', '모험'] |
| 1397 | 82473 | 캐리비안의 해적: 죽은 자는 말이 없다 | Pirates of the Caribbean: Dead Men Tell No Tales | 2017 | 8.28 | 12세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['액션', '모험', '코미디', '판타지'] |
| 1398 | 31606 | 킬러들의 수다 | Guns & Talks | 2001 | 8.28 | 15세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['액션', '드라마', '코미디'] |
| 1399 | 51082 | 7급 공무원 | 7th Grade Civil Servant | 2009 | 8.28 | 12세 관람가 | https://movie.naver.com/movie/bi/mi/basic.nhn?... | ['액션', '코미디'] |
1400 rows × 8 columns
poster_total
array([[[[ 0, 100, 116],
[ 0, 100, 116],
[ 0, 100, 117],
...,
[ 1, 115, 141],
[ 0, 116, 141],
[ 1, 117, 142]],
[[ 0, 100, 116],
[ 0, 100, 115],
[ 0, 101, 118],
...,
[ 0, 115, 141],
[ 0, 116, 141],
[ 0, 116, 141]],
[[ 0, 101, 116],
[ 0, 101, 116],
[ 0, 101, 118],
...,
[ 1, 117, 142],
[ 1, 117, 142],
[ 1, 117, 142]],
...,
[[ 0, 106, 130],
[ 1, 107, 131],
[ 0, 108, 132],
...,
[ 0, 105, 127],
[ 0, 105, 127],
[ 0, 105, 127]],
[[ 0, 106, 130],
[ 0, 106, 130],
[ 1, 107, 131],
...,
[ 1, 104, 127],
[ 1, 104, 126],
[ 1, 104, 127]],
[[ 0, 106, 130],
[ 0, 106, 130],
[ 1, 107, 131],
...,
[ 1, 104, 127],
[ 1, 104, 127],
[ 2, 104, 127]]],
[[[227, 221, 218],
[231, 231, 227],
[237, 238, 239],
...,
[ 87, 56, 93],
[ 89, 58, 94],
[ 89, 57, 92]],
[[231, 229, 226],
[232, 231, 227],
[237, 237, 236],
...,
[ 89, 58, 95],
[ 88, 58, 93],
[ 89, 58, 94]],
[[236, 235, 233],
[236, 236, 235],
[237, 237, 236],
...,
[ 88, 58, 96],
[ 88, 58, 95],
[ 89, 58, 94]],
...,
[[104, 129, 151],
[105, 128, 150],
[106, 130, 152],
...,
[125, 125, 84],
[123, 124, 94],
[ 97, 104, 102]],
[[103, 129, 151],
[102, 128, 150],
[104, 129, 152],
...,
[115, 118, 87],
[111, 115, 91],
[ 87, 96, 99]],
[[100, 129, 149],
[100, 128, 149],
[103, 128, 150],
...,
[106, 112, 85],
[ 97, 106, 86],
[ 79, 85, 93]]],
[[[242, 225, 217],
[241, 223, 213],
[241, 222, 212],
...,
[250, 250, 231],
[253, 252, 248],
[253, 253, 252]],
[[252, 252, 252],
[255, 255, 255],
[255, 255, 255],
...,
[255, 255, 255],
[255, 255, 255],
[255, 255, 255]],
[[247, 234, 229],
[255, 255, 255],
[255, 255, 255],
...,
[255, 255, 255],
[255, 255, 255],
[255, 255, 255]],
...,
[[100, 97, 86],
[ 98, 95, 85],
[ 96, 95, 85],
...,
[114, 102, 86],
[111, 98, 79],
[114, 97, 77]],
[[100, 99, 90],
[ 89, 88, 79],
[ 76, 73, 65],
...,
[114, 100, 81],
[118, 102, 83],
[119, 100, 80]],
[[ 90, 89, 83],
[ 89, 86, 80],
[ 88, 83, 76],
...,
[102, 90, 73],
[104, 92, 76],
[103, 88, 74]]],
...,
[[[107, 147, 155],
[106, 146, 153],
[107, 146, 154],
...,
[214, 222, 204],
[214, 223, 206],
[213, 222, 204]],
[[106, 146, 153],
[106, 147, 154],
[105, 147, 153],
...,
[215, 222, 204],
[214, 222, 206],
[216, 223, 204]],
[[106, 146, 154],
[107, 147, 154],
[104, 146, 152],
...,
[215, 222, 204],
[215, 222, 205],
[210, 220, 204]],
...,
[[ 16, 24, 34],
[ 16, 23, 32],
[ 16, 23, 32],
...,
[ 16, 21, 28],
[ 16, 20, 29],
[ 16, 20, 29]],
[[ 18, 23, 34],
[ 17, 23, 34],
[ 17, 23, 32],
...,
[ 17, 21, 30],
[ 14, 20, 30],
[ 15, 20, 30]],
[[ 18, 22, 33],
[ 18, 23, 35],
[ 17, 22, 33],
...,
[ 20, 22, 30],
[ 45, 29, 30],
[ 28, 22, 29]]],
[[[237, 254, 248],
[250, 250, 250],
[255, 249, 253],
...,
[252, 250, 250],
[254, 250, 250],
[254, 252, 254]],
[[242, 254, 249],
[252, 251, 251],
[244, 243, 245],
...,
[236, 244, 239],
[253, 253, 253],
[254, 252, 255]],
[[248, 253, 251],
[255, 253, 254],
[226, 232, 232],
...,
[163, 172, 164],
[238, 239, 241],
[254, 253, 255]],
...,
[[252, 252, 252],
[255, 255, 255],
[188, 188, 188],
...,
[113, 113, 113],
[220, 220, 220],
[255, 255, 255]],
[[251, 251, 251],
[251, 251, 251],
[249, 249, 249],
...,
[152, 152, 152],
[225, 225, 225],
[255, 255, 255]],
[[252, 252, 252],
[253, 253, 253],
[254, 254, 254],
...,
[252, 252, 252],
[252, 252, 252],
[255, 255, 255]]],
[[[255, 255, 255],
[255, 255, 255],
[255, 255, 255],
...,
[179, 47, 35],
[179, 47, 35],
[179, 47, 35]],
[[255, 255, 255],
[255, 255, 255],
[255, 255, 255],
...,
[180, 48, 33],
[179, 47, 33],
[179, 47, 33]],
[[255, 255, 255],
[255, 255, 255],
[255, 255, 255],
...,
[181, 49, 32],
[181, 49, 35],
[181, 49, 35]],
...,
[[ 96, 4, 8],
[ 58, 2, 1],
[ 15, 1, 0],
...,
[136, 27, 33],
[138, 27, 33],
[138, 27, 33]],
[[ 82, 4, 4],
[ 32, 1, 2],
[ 12, 2, 3],
...,
[136, 27, 32],
[138, 27, 33],
[138, 27, 33]],
[[ 52, 1, 3],
[ 10, 1, 2],
[ 38, 4, 6],
...,
[138, 27, 33],
[137, 26, 32],
[138, 27, 33]]]], dtype=uint8)
연도 확인하기
set(total_df['year'])
{'...',
'1936',
'1939',
'1940',
'1942',
'1953',
'1954',
'1959',
'1960',
'1965',
'1968',
'1972',
'1973',
'1974',
'1975',
'1976',
'1978',
'1979',
'1980',
'1981',
'1983',
'1984',
'1985',
'1986',
'1987',
'1988',
'1989',
'1990',
'1991',
'1992',
'1993',
'1994',
'1995',
'1996',
'1997',
'1998',
'1999',
'19세 관람가 도움말',
'2000',
'2001',
'2002',
'2003',
'2004',
'2005',
'2006',
'2007',
'2008',
'2009',
'201...',
'2010',
'2011',
'2012',
'2013',
'2014',
'2015',
'2016',
'2017',
'2018',
'2019',
'2020'}
이상치 제거
new_df = total_df.loc[(total_df['year'] != '...') & (total_df['year'] != '19세 관람가 도움말') &
(total_df['year'] != '201...')]
new_df['year'].value_counts()
2007 80
2017 72
2016 72
2006 71
2018 68
2004 67
2015 67
2014 64
2009 63
2012 63
2013 61
2010 60
2008 54
2003 53
2011 52
2019 49
2005 48
2002 36
2001 34
2000 29
1999 24
1994 22
1998 19
1995 18
1993 17
1997 17
1992 16
1996 13
1991 9
1989 8
1988 6
1987 6
1990 6
1986 5
2020 4
1984 3
1972 2
1979 2
1975 2
1981 2
1976 2
1973 2
1985 1
1959 1
1960 1
1980 1
1978 1
1954 1
1942 1
1974 1
1968 1
1940 1
1936 1
1983 1
1953 1
1939 1
1965 1
Name: year, dtype: int64
string to int
new_df['year'] = new_df['year'].astype(int)
/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:1: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
"""Entry point for launching an IPython kernel.
연대로 구분하기
def era(year):
# 2010년대 이후
if 2010 <= year:
return 0
# 2000년대
elif 2000 <= year < 2010:
return 1
# 2000년대 이전
else:
return 2
new_df['era'] = new_df['year'].apply(era)
/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:1: SettingWithCopyWarning:
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead
See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
"""Entry point for launching an IPython kernel.
new_df['era'].value_counts()
0 632
1 535
2 216
Name: era, dtype: int64
onehot vector로 만들기
def to_onehot(sequences, dimension=3):
results = np.zeros((len(sequences), dimension))
for i, sequence in enumerate(sequences):
results[i, sequence] = 1.
return results
label_vector = to_onehot(new_df['era'])
label_vector
array([[1., 0., 0.],
[1., 0., 0.],
[1., 0., 0.],
...,
[1., 0., 0.],
[0., 1., 0.],
[0., 1., 0.]])
len(new_df)
1383
new_df.index
Int64Index([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
...
1390, 1391, 1392, 1393, 1394, 1395, 1396, 1397, 1398, 1399],
dtype='int64', length=1383)
inputs scaling
poster_scaled = poster_total[new_df.index].astype('float32') / 255
3. Modeling
train / test 나누기
train 60%, validation 20%, test 20%로 데이터를 나누면
총 1383개의 데이터 중 829개가 train, 277개가 validation, 277개가 test가 됨
# test 데이터로 20%를 뽑기 위한 랜덤 난수 생성
np.random.seed(303)
shuffled_index = np.random.permutation(1383)
shuffled_index
array([ 176, 420, 322, ..., 1153, 530, 571])
# train: 0 ~ 1105번 index (1106개)
# test: 1106 ~ 1383번 index (277개)
x_train = poster_scaled[shuffled_index[:1106]]
x_test = poster_scaled[shuffled_index[1106:]]
y_train = label_vector[shuffled_index[:1106]]
y_test = label_vector[shuffled_index[1106:]]
train data shuffling
np.random.seed(8989)
s_index = np.random.permutation(1106)
x_train = x_train[s_index]
y_train = y_train[s_index]
x_train.shape
(1106, 256, 256, 3)
y_train.shape
(1106, 3)
a. Basic CNN
def build_model():
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', padding='same',
input_shape=(256, 256, 3)))
model.add(layers.Conv2D(32, (3, 3), activation='relu', padding='same'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu', padding='same'))
model.add(layers.Conv2D(64, (3, 3), activation='relu', padding='same'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu', padding='same'))
model.add(layers.Conv2D(128, (3, 3), activation='relu', padding='same'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(3, activation='softmax'))
model.compile(optimizer=optimizers.RMSprop(lr=0.001),
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
model = build_model()
model.name = 'default_model'
model.summary()
Model: "default_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d_1 (Conv2D) (None, 256, 256, 32) 896
_________________________________________________________________
conv2d_2 (Conv2D) (None, 256, 256, 32) 9248
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 128, 128, 32) 0
_________________________________________________________________
conv2d_3 (Conv2D) (None, 128, 128, 64) 18496
_________________________________________________________________
conv2d_4 (Conv2D) (None, 128, 128, 64) 36928
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 64, 64, 64) 0
_________________________________________________________________
conv2d_5 (Conv2D) (None, 64, 64, 128) 73856
_________________________________________________________________
conv2d_6 (Conv2D) (None, 64, 64, 128) 147584
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 32, 32, 128) 0
_________________________________________________________________
flatten_1 (Flatten) (None, 131072) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 67109376
_________________________________________________________________
dense_2 (Dense) (None, 3) 1539
=================================================================
Total params: 67,397,923
Trainable params: 67,397,923
Non-trainable params: 0
_________________________________________________________________
def model_train(model, num_epochs):
filepath = '/content/drive/Shared drives/personal/weights/' + model.name + '.{epoch:02d}-{val_accuracy:.4f}.hdf5'
checkpoint = ModelCheckpoint(filepath, monitor='val_accuracy', verbose=0, save_weights_only=True, save_best_only=True, mode='auto', period=1)
model_history = model.fit(x_train, y_train, epochs=num_epochs, batch_size = 16, verbose=1, callbacks=[checkpoint], validation_split=0.25, shuffle=True)
return model.history
history = model_train(model, 20)
Train on 829 samples, validate on 277 samples
Epoch 1/20
829/829 [==============================] - 11s 13ms/step - loss: 3.4242 - accuracy: 0.4282 - val_loss: 1.0881 - val_accuracy: 0.3574
Epoch 2/20
829/829 [==============================] - 4s 4ms/step - loss: 1.0365 - accuracy: 0.4632 - val_loss: 1.0050 - val_accuracy: 0.4874
Epoch 3/20
829/829 [==============================] - 4s 4ms/step - loss: 1.0529 - accuracy: 0.4415 - val_loss: 1.0087 - val_accuracy: 0.4765
Epoch 4/20
829/829 [==============================] - 4s 4ms/step - loss: 1.1031 - accuracy: 0.4572 - val_loss: 1.0188 - val_accuracy: 0.4549
Epoch 5/20
829/829 [==============================] - 4s 4ms/step - loss: 1.0396 - accuracy: 0.5139 - val_loss: 1.1308 - val_accuracy: 0.4910
Epoch 6/20
829/829 [==============================] - 4s 4ms/step - loss: 0.9136 - accuracy: 0.5875 - val_loss: 1.0257 - val_accuracy: 0.5090
Epoch 7/20
829/829 [==============================] - 4s 4ms/step - loss: 0.7788 - accuracy: 0.6381 - val_loss: 1.0869 - val_accuracy: 0.4188
Epoch 8/20
829/829 [==============================] - 4s 4ms/step - loss: 0.6363 - accuracy: 0.7419 - val_loss: 1.2535 - val_accuracy: 0.4477
Epoch 9/20
829/829 [==============================] - 4s 4ms/step - loss: 0.3825 - accuracy: 0.8480 - val_loss: 2.6986 - val_accuracy: 0.4296
Epoch 10/20
829/829 [==============================] - 4s 4ms/step - loss: 0.2134 - accuracy: 0.9192 - val_loss: 3.7182 - val_accuracy: 0.4332
Epoch 11/20
829/829 [==============================] - 4s 4ms/step - loss: 0.1467 - accuracy: 0.9554 - val_loss: 3.2280 - val_accuracy: 0.4440
Epoch 12/20
829/829 [==============================] - 4s 4ms/step - loss: 0.1742 - accuracy: 0.9831 - val_loss: 5.8329 - val_accuracy: 0.4585
Epoch 13/20
829/829 [==============================] - 4s 4ms/step - loss: 0.1002 - accuracy: 0.9843 - val_loss: 5.0345 - val_accuracy: 0.4874
Epoch 14/20
829/829 [==============================] - 4s 4ms/step - loss: 0.0016 - accuracy: 1.0000 - val_loss: 8.9527 - val_accuracy: 0.4440
Epoch 15/20
829/829 [==============================] - 4s 4ms/step - loss: 2.7524e-05 - accuracy: 1.0000 - val_loss: 11.7579 - val_accuracy: 0.4910
Epoch 16/20
829/829 [==============================] - 4s 4ms/step - loss: 2.6138e-06 - accuracy: 1.0000 - val_loss: 14.4503 - val_accuracy: 0.4729
Epoch 17/20
829/829 [==============================] - 4s 4ms/step - loss: 1.7256e-08 - accuracy: 1.0000 - val_loss: 14.9324 - val_accuracy: 0.4874
Epoch 18/20
829/829 [==============================] - 4s 4ms/step - loss: 6.1833e-09 - accuracy: 1.0000 - val_loss: 16.6347 - val_accuracy: 0.4657
Epoch 19/20
829/829 [==============================] - 4s 4ms/step - loss: 2.8760e-10 - accuracy: 1.0000 - val_loss: 16.6152 - val_accuracy: 0.4693
Epoch 20/20
829/829 [==============================] - 4s 4ms/step - loss: 1.4380e-10 - accuracy: 1.0000 - val_loss: 16.5649 - val_accuracy: 0.4621
history_dict = history.history
loss = history_dict['loss']
val_loss = history_dict['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
acc = history_dict['accuracy']
val_acc = history_dict['val_accuracy']
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation acc')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
model.load_weights('/content/drive/Shared drives/personal/weights/default_model.06-0.5090.hdf5')
model.evaluate(x_test, y_test)
277/277 [==============================] - 1s 3ms/step
[0.9569806161770321, 0.505415141582489]
b. ResNet50
def build_resnet():
# keras 기본 모델 이용
base_model = resnet50.ResNet50(weights=None, include_top=False, input_shape=(256, 256 ,3))
output_model = base_model.output
flatten_model = layers.Flatten()(output_model)
dense_model = layers.Dense(512, activation='relu')(flatten_model)
final_model = layers.Dense(3, activation='softmax')(dense_model)
model = models.Model(inputs=base_model.input, outputs=final_model)
model.compile(optimizer=optimizers.RMSprop(lr=0.001),
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
model = build_resnet()
/usr/local/lib/python3.6/dist-packages/keras_applications/resnet50.py:265: UserWarning: The output shape of `ResNet50(include_top=False)` has been changed since Keras 2.2.0.
warnings.warn('The output shape of `ResNet50(include_top=False)` '
model.name = 'resnet_model'
model.summary()
Model: "resnet_model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_8 (InputLayer) (None, 256, 256, 3) 0
__________________________________________________________________________________________________
conv1_pad (ZeroPadding2D) (None, 262, 262, 3) 0 input_8[0][0]
__________________________________________________________________________________________________
conv1 (Conv2D) (None, 128, 128, 64) 9472 conv1_pad[0][0]
__________________________________________________________________________________________________
bn_conv1 (BatchNormalization) (None, 128, 128, 64) 256 conv1[0][0]
__________________________________________________________________________________________________
activation_246 (Activation) (None, 128, 128, 64) 0 bn_conv1[0][0]
__________________________________________________________________________________________________
pool1_pad (ZeroPadding2D) (None, 130, 130, 64) 0 activation_246[0][0]
__________________________________________________________________________________________________
max_pooling2d_17 (MaxPooling2D) (None, 64, 64, 64) 0 pool1_pad[0][0]
__________________________________________________________________________________________________
res2a_branch2a (Conv2D) (None, 64, 64, 64) 4160 max_pooling2d_17[0][0]
__________________________________________________________________________________________________
bn2a_branch2a (BatchNormalizati (None, 64, 64, 64) 256 res2a_branch2a[0][0]
__________________________________________________________________________________________________
activation_247 (Activation) (None, 64, 64, 64) 0 bn2a_branch2a[0][0]
__________________________________________________________________________________________________
res2a_branch2b (Conv2D) (None, 64, 64, 64) 36928 activation_247[0][0]
__________________________________________________________________________________________________
bn2a_branch2b (BatchNormalizati (None, 64, 64, 64) 256 res2a_branch2b[0][0]
__________________________________________________________________________________________________
activation_248 (Activation) (None, 64, 64, 64) 0 bn2a_branch2b[0][0]
__________________________________________________________________________________________________
res2a_branch2c (Conv2D) (None, 64, 64, 256) 16640 activation_248[0][0]
__________________________________________________________________________________________________
res2a_branch1 (Conv2D) (None, 64, 64, 256) 16640 max_pooling2d_17[0][0]
__________________________________________________________________________________________________
bn2a_branch2c (BatchNormalizati (None, 64, 64, 256) 1024 res2a_branch2c[0][0]
__________________________________________________________________________________________________
bn2a_branch1 (BatchNormalizatio (None, 64, 64, 256) 1024 res2a_branch1[0][0]
__________________________________________________________________________________________________
add_81 (Add) (None, 64, 64, 256) 0 bn2a_branch2c[0][0]
bn2a_branch1[0][0]
__________________________________________________________________________________________________
activation_249 (Activation) (None, 64, 64, 256) 0 add_81[0][0]
__________________________________________________________________________________________________
res2b_branch2a (Conv2D) (None, 64, 64, 64) 16448 activation_249[0][0]
__________________________________________________________________________________________________
bn2b_branch2a (BatchNormalizati (None, 64, 64, 64) 256 res2b_branch2a[0][0]
__________________________________________________________________________________________________
activation_250 (Activation) (None, 64, 64, 64) 0 bn2b_branch2a[0][0]
__________________________________________________________________________________________________
res2b_branch2b (Conv2D) (None, 64, 64, 64) 36928 activation_250[0][0]
__________________________________________________________________________________________________
bn2b_branch2b (BatchNormalizati (None, 64, 64, 64) 256 res2b_branch2b[0][0]
__________________________________________________________________________________________________
activation_251 (Activation) (None, 64, 64, 64) 0 bn2b_branch2b[0][0]
__________________________________________________________________________________________________
res2b_branch2c (Conv2D) (None, 64, 64, 256) 16640 activation_251[0][0]
__________________________________________________________________________________________________
bn2b_branch2c (BatchNormalizati (None, 64, 64, 256) 1024 res2b_branch2c[0][0]
__________________________________________________________________________________________________
add_82 (Add) (None, 64, 64, 256) 0 bn2b_branch2c[0][0]
activation_249[0][0]
__________________________________________________________________________________________________
activation_252 (Activation) (None, 64, 64, 256) 0 add_82[0][0]
__________________________________________________________________________________________________
res2c_branch2a (Conv2D) (None, 64, 64, 64) 16448 activation_252[0][0]
__________________________________________________________________________________________________
bn2c_branch2a (BatchNormalizati (None, 64, 64, 64) 256 res2c_branch2a[0][0]
__________________________________________________________________________________________________
activation_253 (Activation) (None, 64, 64, 64) 0 bn2c_branch2a[0][0]
__________________________________________________________________________________________________
res2c_branch2b (Conv2D) (None, 64, 64, 64) 36928 activation_253[0][0]
__________________________________________________________________________________________________
bn2c_branch2b (BatchNormalizati (None, 64, 64, 64) 256 res2c_branch2b[0][0]
__________________________________________________________________________________________________
activation_254 (Activation) (None, 64, 64, 64) 0 bn2c_branch2b[0][0]
__________________________________________________________________________________________________
res2c_branch2c (Conv2D) (None, 64, 64, 256) 16640 activation_254[0][0]
__________________________________________________________________________________________________
bn2c_branch2c (BatchNormalizati (None, 64, 64, 256) 1024 res2c_branch2c[0][0]
__________________________________________________________________________________________________
add_83 (Add) (None, 64, 64, 256) 0 bn2c_branch2c[0][0]
activation_252[0][0]
__________________________________________________________________________________________________
activation_255 (Activation) (None, 64, 64, 256) 0 add_83[0][0]
__________________________________________________________________________________________________
res3a_branch2a (Conv2D) (None, 32, 32, 128) 32896 activation_255[0][0]
__________________________________________________________________________________________________
bn3a_branch2a (BatchNormalizati (None, 32, 32, 128) 512 res3a_branch2a[0][0]
__________________________________________________________________________________________________
activation_256 (Activation) (None, 32, 32, 128) 0 bn3a_branch2a[0][0]
__________________________________________________________________________________________________
res3a_branch2b (Conv2D) (None, 32, 32, 128) 147584 activation_256[0][0]
__________________________________________________________________________________________________
bn3a_branch2b (BatchNormalizati (None, 32, 32, 128) 512 res3a_branch2b[0][0]
__________________________________________________________________________________________________
activation_257 (Activation) (None, 32, 32, 128) 0 bn3a_branch2b[0][0]
__________________________________________________________________________________________________
res3a_branch2c (Conv2D) (None, 32, 32, 512) 66048 activation_257[0][0]
__________________________________________________________________________________________________
res3a_branch1 (Conv2D) (None, 32, 32, 512) 131584 activation_255[0][0]
__________________________________________________________________________________________________
bn3a_branch2c (BatchNormalizati (None, 32, 32, 512) 2048 res3a_branch2c[0][0]
__________________________________________________________________________________________________
bn3a_branch1 (BatchNormalizatio (None, 32, 32, 512) 2048 res3a_branch1[0][0]
__________________________________________________________________________________________________
add_84 (Add) (None, 32, 32, 512) 0 bn3a_branch2c[0][0]
bn3a_branch1[0][0]
__________________________________________________________________________________________________
activation_258 (Activation) (None, 32, 32, 512) 0 add_84[0][0]
__________________________________________________________________________________________________
res3b_branch2a (Conv2D) (None, 32, 32, 128) 65664 activation_258[0][0]
__________________________________________________________________________________________________
bn3b_branch2a (BatchNormalizati (None, 32, 32, 128) 512 res3b_branch2a[0][0]
__________________________________________________________________________________________________
activation_259 (Activation) (None, 32, 32, 128) 0 bn3b_branch2a[0][0]
__________________________________________________________________________________________________
res3b_branch2b (Conv2D) (None, 32, 32, 128) 147584 activation_259[0][0]
__________________________________________________________________________________________________
bn3b_branch2b (BatchNormalizati (None, 32, 32, 128) 512 res3b_branch2b[0][0]
__________________________________________________________________________________________________
activation_260 (Activation) (None, 32, 32, 128) 0 bn3b_branch2b[0][0]
__________________________________________________________________________________________________
res3b_branch2c (Conv2D) (None, 32, 32, 512) 66048 activation_260[0][0]
__________________________________________________________________________________________________
bn3b_branch2c (BatchNormalizati (None, 32, 32, 512) 2048 res3b_branch2c[0][0]
__________________________________________________________________________________________________
add_85 (Add) (None, 32, 32, 512) 0 bn3b_branch2c[0][0]
activation_258[0][0]
__________________________________________________________________________________________________
activation_261 (Activation) (None, 32, 32, 512) 0 add_85[0][0]
__________________________________________________________________________________________________
res3c_branch2a (Conv2D) (None, 32, 32, 128) 65664 activation_261[0][0]
__________________________________________________________________________________________________
bn3c_branch2a (BatchNormalizati (None, 32, 32, 128) 512 res3c_branch2a[0][0]
__________________________________________________________________________________________________
activation_262 (Activation) (None, 32, 32, 128) 0 bn3c_branch2a[0][0]
__________________________________________________________________________________________________
res3c_branch2b (Conv2D) (None, 32, 32, 128) 147584 activation_262[0][0]
__________________________________________________________________________________________________
bn3c_branch2b (BatchNormalizati (None, 32, 32, 128) 512 res3c_branch2b[0][0]
__________________________________________________________________________________________________
activation_263 (Activation) (None, 32, 32, 128) 0 bn3c_branch2b[0][0]
__________________________________________________________________________________________________
res3c_branch2c (Conv2D) (None, 32, 32, 512) 66048 activation_263[0][0]
__________________________________________________________________________________________________
bn3c_branch2c (BatchNormalizati (None, 32, 32, 512) 2048 res3c_branch2c[0][0]
__________________________________________________________________________________________________
add_86 (Add) (None, 32, 32, 512) 0 bn3c_branch2c[0][0]
activation_261[0][0]
__________________________________________________________________________________________________
activation_264 (Activation) (None, 32, 32, 512) 0 add_86[0][0]
__________________________________________________________________________________________________
res3d_branch2a (Conv2D) (None, 32, 32, 128) 65664 activation_264[0][0]
__________________________________________________________________________________________________
bn3d_branch2a (BatchNormalizati (None, 32, 32, 128) 512 res3d_branch2a[0][0]
__________________________________________________________________________________________________
activation_265 (Activation) (None, 32, 32, 128) 0 bn3d_branch2a[0][0]
__________________________________________________________________________________________________
res3d_branch2b (Conv2D) (None, 32, 32, 128) 147584 activation_265[0][0]
__________________________________________________________________________________________________
bn3d_branch2b (BatchNormalizati (None, 32, 32, 128) 512 res3d_branch2b[0][0]
__________________________________________________________________________________________________
activation_266 (Activation) (None, 32, 32, 128) 0 bn3d_branch2b[0][0]
__________________________________________________________________________________________________
res3d_branch2c (Conv2D) (None, 32, 32, 512) 66048 activation_266[0][0]
__________________________________________________________________________________________________
bn3d_branch2c (BatchNormalizati (None, 32, 32, 512) 2048 res3d_branch2c[0][0]
__________________________________________________________________________________________________
add_87 (Add) (None, 32, 32, 512) 0 bn3d_branch2c[0][0]
activation_264[0][0]
__________________________________________________________________________________________________
activation_267 (Activation) (None, 32, 32, 512) 0 add_87[0][0]
__________________________________________________________________________________________________
res4a_branch2a (Conv2D) (None, 16, 16, 256) 131328 activation_267[0][0]
__________________________________________________________________________________________________
bn4a_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4a_branch2a[0][0]
__________________________________________________________________________________________________
activation_268 (Activation) (None, 16, 16, 256) 0 bn4a_branch2a[0][0]
__________________________________________________________________________________________________
res4a_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_268[0][0]
__________________________________________________________________________________________________
bn4a_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4a_branch2b[0][0]
__________________________________________________________________________________________________
activation_269 (Activation) (None, 16, 16, 256) 0 bn4a_branch2b[0][0]
__________________________________________________________________________________________________
res4a_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_269[0][0]
__________________________________________________________________________________________________
res4a_branch1 (Conv2D) (None, 16, 16, 1024) 525312 activation_267[0][0]
__________________________________________________________________________________________________
bn4a_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4a_branch2c[0][0]
__________________________________________________________________________________________________
bn4a_branch1 (BatchNormalizatio (None, 16, 16, 1024) 4096 res4a_branch1[0][0]
__________________________________________________________________________________________________
add_88 (Add) (None, 16, 16, 1024) 0 bn4a_branch2c[0][0]
bn4a_branch1[0][0]
__________________________________________________________________________________________________
activation_270 (Activation) (None, 16, 16, 1024) 0 add_88[0][0]
__________________________________________________________________________________________________
res4b_branch2a (Conv2D) (None, 16, 16, 256) 262400 activation_270[0][0]
__________________________________________________________________________________________________
bn4b_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4b_branch2a[0][0]
__________________________________________________________________________________________________
activation_271 (Activation) (None, 16, 16, 256) 0 bn4b_branch2a[0][0]
__________________________________________________________________________________________________
res4b_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_271[0][0]
__________________________________________________________________________________________________
bn4b_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4b_branch2b[0][0]
__________________________________________________________________________________________________
activation_272 (Activation) (None, 16, 16, 256) 0 bn4b_branch2b[0][0]
__________________________________________________________________________________________________
res4b_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_272[0][0]
__________________________________________________________________________________________________
bn4b_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4b_branch2c[0][0]
__________________________________________________________________________________________________
add_89 (Add) (None, 16, 16, 1024) 0 bn4b_branch2c[0][0]
activation_270[0][0]
__________________________________________________________________________________________________
activation_273 (Activation) (None, 16, 16, 1024) 0 add_89[0][0]
__________________________________________________________________________________________________
res4c_branch2a (Conv2D) (None, 16, 16, 256) 262400 activation_273[0][0]
__________________________________________________________________________________________________
bn4c_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4c_branch2a[0][0]
__________________________________________________________________________________________________
activation_274 (Activation) (None, 16, 16, 256) 0 bn4c_branch2a[0][0]
__________________________________________________________________________________________________
res4c_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_274[0][0]
__________________________________________________________________________________________________
bn4c_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4c_branch2b[0][0]
__________________________________________________________________________________________________
activation_275 (Activation) (None, 16, 16, 256) 0 bn4c_branch2b[0][0]
__________________________________________________________________________________________________
res4c_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_275[0][0]
__________________________________________________________________________________________________
bn4c_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4c_branch2c[0][0]
__________________________________________________________________________________________________
add_90 (Add) (None, 16, 16, 1024) 0 bn4c_branch2c[0][0]
activation_273[0][0]
__________________________________________________________________________________________________
activation_276 (Activation) (None, 16, 16, 1024) 0 add_90[0][0]
__________________________________________________________________________________________________
res4d_branch2a (Conv2D) (None, 16, 16, 256) 262400 activation_276[0][0]
__________________________________________________________________________________________________
bn4d_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4d_branch2a[0][0]
__________________________________________________________________________________________________
activation_277 (Activation) (None, 16, 16, 256) 0 bn4d_branch2a[0][0]
__________________________________________________________________________________________________
res4d_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_277[0][0]
__________________________________________________________________________________________________
bn4d_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4d_branch2b[0][0]
__________________________________________________________________________________________________
activation_278 (Activation) (None, 16, 16, 256) 0 bn4d_branch2b[0][0]
__________________________________________________________________________________________________
res4d_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_278[0][0]
__________________________________________________________________________________________________
bn4d_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4d_branch2c[0][0]
__________________________________________________________________________________________________
add_91 (Add) (None, 16, 16, 1024) 0 bn4d_branch2c[0][0]
activation_276[0][0]
__________________________________________________________________________________________________
activation_279 (Activation) (None, 16, 16, 1024) 0 add_91[0][0]
__________________________________________________________________________________________________
res4e_branch2a (Conv2D) (None, 16, 16, 256) 262400 activation_279[0][0]
__________________________________________________________________________________________________
bn4e_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4e_branch2a[0][0]
__________________________________________________________________________________________________
activation_280 (Activation) (None, 16, 16, 256) 0 bn4e_branch2a[0][0]
__________________________________________________________________________________________________
res4e_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_280[0][0]
__________________________________________________________________________________________________
bn4e_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4e_branch2b[0][0]
__________________________________________________________________________________________________
activation_281 (Activation) (None, 16, 16, 256) 0 bn4e_branch2b[0][0]
__________________________________________________________________________________________________
res4e_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_281[0][0]
__________________________________________________________________________________________________
bn4e_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4e_branch2c[0][0]
__________________________________________________________________________________________________
add_92 (Add) (None, 16, 16, 1024) 0 bn4e_branch2c[0][0]
activation_279[0][0]
__________________________________________________________________________________________________
activation_282 (Activation) (None, 16, 16, 1024) 0 add_92[0][0]
__________________________________________________________________________________________________
res4f_branch2a (Conv2D) (None, 16, 16, 256) 262400 activation_282[0][0]
__________________________________________________________________________________________________
bn4f_branch2a (BatchNormalizati (None, 16, 16, 256) 1024 res4f_branch2a[0][0]
__________________________________________________________________________________________________
activation_283 (Activation) (None, 16, 16, 256) 0 bn4f_branch2a[0][0]
__________________________________________________________________________________________________
res4f_branch2b (Conv2D) (None, 16, 16, 256) 590080 activation_283[0][0]
__________________________________________________________________________________________________
bn4f_branch2b (BatchNormalizati (None, 16, 16, 256) 1024 res4f_branch2b[0][0]
__________________________________________________________________________________________________
activation_284 (Activation) (None, 16, 16, 256) 0 bn4f_branch2b[0][0]
__________________________________________________________________________________________________
res4f_branch2c (Conv2D) (None, 16, 16, 1024) 263168 activation_284[0][0]
__________________________________________________________________________________________________
bn4f_branch2c (BatchNormalizati (None, 16, 16, 1024) 4096 res4f_branch2c[0][0]
__________________________________________________________________________________________________
add_93 (Add) (None, 16, 16, 1024) 0 bn4f_branch2c[0][0]
activation_282[0][0]
__________________________________________________________________________________________________
activation_285 (Activation) (None, 16, 16, 1024) 0 add_93[0][0]
__________________________________________________________________________________________________
res5a_branch2a (Conv2D) (None, 8, 8, 512) 524800 activation_285[0][0]
__________________________________________________________________________________________________
bn5a_branch2a (BatchNormalizati (None, 8, 8, 512) 2048 res5a_branch2a[0][0]
__________________________________________________________________________________________________
activation_286 (Activation) (None, 8, 8, 512) 0 bn5a_branch2a[0][0]
__________________________________________________________________________________________________
res5a_branch2b (Conv2D) (None, 8, 8, 512) 2359808 activation_286[0][0]
__________________________________________________________________________________________________
bn5a_branch2b (BatchNormalizati (None, 8, 8, 512) 2048 res5a_branch2b[0][0]
__________________________________________________________________________________________________
activation_287 (Activation) (None, 8, 8, 512) 0 bn5a_branch2b[0][0]
__________________________________________________________________________________________________
res5a_branch2c (Conv2D) (None, 8, 8, 2048) 1050624 activation_287[0][0]
__________________________________________________________________________________________________
res5a_branch1 (Conv2D) (None, 8, 8, 2048) 2099200 activation_285[0][0]
__________________________________________________________________________________________________
bn5a_branch2c (BatchNormalizati (None, 8, 8, 2048) 8192 res5a_branch2c[0][0]
__________________________________________________________________________________________________
bn5a_branch1 (BatchNormalizatio (None, 8, 8, 2048) 8192 res5a_branch1[0][0]
__________________________________________________________________________________________________
add_94 (Add) (None, 8, 8, 2048) 0 bn5a_branch2c[0][0]
bn5a_branch1[0][0]
__________________________________________________________________________________________________
activation_288 (Activation) (None, 8, 8, 2048) 0 add_94[0][0]
__________________________________________________________________________________________________
res5b_branch2a (Conv2D) (None, 8, 8, 512) 1049088 activation_288[0][0]
__________________________________________________________________________________________________
bn5b_branch2a (BatchNormalizati (None, 8, 8, 512) 2048 res5b_branch2a[0][0]
__________________________________________________________________________________________________
activation_289 (Activation) (None, 8, 8, 512) 0 bn5b_branch2a[0][0]
__________________________________________________________________________________________________
res5b_branch2b (Conv2D) (None, 8, 8, 512) 2359808 activation_289[0][0]
__________________________________________________________________________________________________
bn5b_branch2b (BatchNormalizati (None, 8, 8, 512) 2048 res5b_branch2b[0][0]
__________________________________________________________________________________________________
activation_290 (Activation) (None, 8, 8, 512) 0 bn5b_branch2b[0][0]
__________________________________________________________________________________________________
res5b_branch2c (Conv2D) (None, 8, 8, 2048) 1050624 activation_290[0][0]
__________________________________________________________________________________________________
bn5b_branch2c (BatchNormalizati (None, 8, 8, 2048) 8192 res5b_branch2c[0][0]
__________________________________________________________________________________________________
add_95 (Add) (None, 8, 8, 2048) 0 bn5b_branch2c[0][0]
activation_288[0][0]
__________________________________________________________________________________________________
activation_291 (Activation) (None, 8, 8, 2048) 0 add_95[0][0]
__________________________________________________________________________________________________
res5c_branch2a (Conv2D) (None, 8, 8, 512) 1049088 activation_291[0][0]
__________________________________________________________________________________________________
bn5c_branch2a (BatchNormalizati (None, 8, 8, 512) 2048 res5c_branch2a[0][0]
__________________________________________________________________________________________________
activation_292 (Activation) (None, 8, 8, 512) 0 bn5c_branch2a[0][0]
__________________________________________________________________________________________________
res5c_branch2b (Conv2D) (None, 8, 8, 512) 2359808 activation_292[0][0]
__________________________________________________________________________________________________
bn5c_branch2b (BatchNormalizati (None, 8, 8, 512) 2048 res5c_branch2b[0][0]
__________________________________________________________________________________________________
activation_293 (Activation) (None, 8, 8, 512) 0 bn5c_branch2b[0][0]
__________________________________________________________________________________________________
res5c_branch2c (Conv2D) (None, 8, 8, 2048) 1050624 activation_293[0][0]
__________________________________________________________________________________________________
bn5c_branch2c (BatchNormalizati (None, 8, 8, 2048) 8192 res5c_branch2c[0][0]
__________________________________________________________________________________________________
add_96 (Add) (None, 8, 8, 2048) 0 bn5c_branch2c[0][0]
activation_291[0][0]
__________________________________________________________________________________________________
activation_294 (Activation) (None, 8, 8, 2048) 0 add_96[0][0]
__________________________________________________________________________________________________
flatten_5 (Flatten) (None, 131072) 0 activation_294[0][0]
__________________________________________________________________________________________________
dense_12 (Dense) (None, 512) 67109376 flatten_5[0][0]
__________________________________________________________________________________________________
dense_13 (Dense) (None, 3) 1539 dense_12[0][0]
==================================================================================================
Total params: 90,698,627
Trainable params: 90,645,507
Non-trainable params: 53,120
__________________________________________________________________________________________________
history = model_train(model, 40)
Train on 829 samples, validate on 277 samples
Epoch 1/40
829/829 [==============================] - 27s 32ms/step - loss: 110.0953 - accuracy: 0.3390 - val_loss: 1355.0505 - val_accuracy: 0.3069
Epoch 2/40
829/829 [==============================] - 15s 18ms/step - loss: 4.3456 - accuracy: 0.4089 - val_loss: 56807.1282 - val_accuracy: 0.3430
Epoch 3/40
829/829 [==============================] - 14s 17ms/step - loss: 1.9580 - accuracy: 0.4101 - val_loss: 776.6751 - val_accuracy: 0.4043
Epoch 4/40
829/829 [==============================] - 14s 17ms/step - loss: 1.2369 - accuracy: 0.4234 - val_loss: 2.1374 - val_accuracy: 0.4513
Epoch 5/40
829/829 [==============================] - 15s 18ms/step - loss: 1.0994 - accuracy: 0.4017 - val_loss: 6.5897 - val_accuracy: 0.2996
Epoch 6/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0461 - accuracy: 0.4343 - val_loss: 40.6376 - val_accuracy: 0.3682
Epoch 7/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0381 - accuracy: 0.4234 - val_loss: 83.7707 - val_accuracy: 0.3827
Epoch 8/40
829/829 [==============================] - 14s 17ms/step - loss: 1.1470 - accuracy: 0.4234 - val_loss: 1.0714 - val_accuracy: 0.4910
Epoch 9/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0752 - accuracy: 0.4524 - val_loss: 1.1472 - val_accuracy: 0.4621
Epoch 10/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0367 - accuracy: 0.4680 - val_loss: 4.5824 - val_accuracy: 0.3574
Epoch 11/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0282 - accuracy: 0.4379 - val_loss: 48.4230 - val_accuracy: 0.4838
Epoch 12/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0280 - accuracy: 0.4258 - val_loss: 2.9461 - val_accuracy: 0.4910
Epoch 13/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0164 - accuracy: 0.4258 - val_loss: 14.7316 - val_accuracy: 0.4765
Epoch 14/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0365 - accuracy: 0.4270 - val_loss: 30.3243 - val_accuracy: 0.4874
Epoch 15/40
829/829 [==============================] - 14s 17ms/step - loss: 1.0087 - accuracy: 0.4210 - val_loss: 2.2417 - val_accuracy: 0.4693
Epoch 16/40
829/829 [==============================] - 14s 17ms/step - loss: 0.9984 - accuracy: 0.4246 - val_loss: 3.3897 - val_accuracy: 0.3755
Epoch 17/40
829/829 [==============================] - 14s 17ms/step - loss: 0.9947 - accuracy: 0.4210 - val_loss: 2.5354 - val_accuracy: 0.4332
Epoch 18/40
829/829 [==============================] - 15s 18ms/step - loss: 0.9816 - accuracy: 0.4294 - val_loss: 2.5836 - val_accuracy: 0.4260
Epoch 19/40
829/829 [==============================] - 15s 18ms/step - loss: 0.9905 - accuracy: 0.4367 - val_loss: 3.7856 - val_accuracy: 0.4946
Epoch 20/40
829/829 [==============================] - 15s 18ms/step - loss: 0.9871 - accuracy: 0.4367 - val_loss: 5.0321 - val_accuracy: 0.3610
Epoch 21/40
829/829 [==============================] - 15s 17ms/step - loss: 0.9815 - accuracy: 0.4511 - val_loss: 1.1294 - val_accuracy: 0.4838
Epoch 22/40
829/829 [==============================] - 15s 17ms/step - loss: 0.9706 - accuracy: 0.4499 - val_loss: 1.0897 - val_accuracy: 0.5126
Epoch 23/40
829/829 [==============================] - 15s 17ms/step - loss: 0.9557 - accuracy: 0.4560 - val_loss: 1.1240 - val_accuracy: 0.5054
Epoch 24/40
829/829 [==============================] - 14s 17ms/step - loss: 0.9380 - accuracy: 0.4511 - val_loss: 2.8845 - val_accuracy: 0.3755
Epoch 25/40
829/829 [==============================] - 15s 18ms/step - loss: 0.9309 - accuracy: 0.4596 - val_loss: 1.5849 - val_accuracy: 0.4765
Epoch 26/40
829/829 [==============================] - 15s 18ms/step - loss: 0.9354 - accuracy: 0.4885 - val_loss: 1.0172 - val_accuracy: 0.5126
Epoch 27/40
829/829 [==============================] - 14s 17ms/step - loss: 0.8974 - accuracy: 0.5139 - val_loss: 1.3205 - val_accuracy: 0.4946
Epoch 28/40
829/829 [==============================] - 14s 17ms/step - loss: 0.8663 - accuracy: 0.5042 - val_loss: 1.4609 - val_accuracy: 0.4657
Epoch 29/40
829/829 [==============================] - 15s 17ms/step - loss: 0.8372 - accuracy: 0.5404 - val_loss: 1.1364 - val_accuracy: 0.4874
Epoch 30/40
829/829 [==============================] - 15s 18ms/step - loss: 0.7965 - accuracy: 0.5344 - val_loss: 2.7651 - val_accuracy: 0.4946
Epoch 31/40
829/829 [==============================] - 15s 18ms/step - loss: 0.7794 - accuracy: 0.5633 - val_loss: 1.0393 - val_accuracy: 0.5379
Epoch 32/40
829/829 [==============================] - 15s 18ms/step - loss: 0.7651 - accuracy: 0.5959 - val_loss: 1.7666 - val_accuracy: 0.4621
Epoch 33/40
829/829 [==============================] - 15s 18ms/step - loss: 0.7173 - accuracy: 0.6671 - val_loss: 2.4147 - val_accuracy: 0.4621
Epoch 34/40
829/829 [==============================] - 15s 18ms/step - loss: 0.6985 - accuracy: 0.6731 - val_loss: 1.9983 - val_accuracy: 0.4729
Epoch 35/40
829/829 [==============================] - 14s 17ms/step - loss: 0.6956 - accuracy: 0.7081 - val_loss: 3.1167 - val_accuracy: 0.4585
Epoch 36/40
829/829 [==============================] - 15s 18ms/step - loss: 0.6369 - accuracy: 0.7165 - val_loss: 1.8084 - val_accuracy: 0.4224
Epoch 37/40
829/829 [==============================] - 15s 18ms/step - loss: 0.5775 - accuracy: 0.7660 - val_loss: 1.4000 - val_accuracy: 0.4729
Epoch 38/40
829/829 [==============================] - 15s 18ms/step - loss: 0.5752 - accuracy: 0.7503 - val_loss: 1.7759 - val_accuracy: 0.4188
Epoch 39/40
829/829 [==============================] - 14s 17ms/step - loss: 0.5381 - accuracy: 0.7853 - val_loss: 2.1087 - val_accuracy: 0.4801
Epoch 40/40
829/829 [==============================] - 15s 18ms/step - loss: 0.4921 - accuracy: 0.7998 - val_loss: 2.5755 - val_accuracy: 0.4946
history_dict = history.history
loss = history_dict['loss']
val_loss = history_dict['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
acc = history_dict['accuracy']
val_acc = history_dict['val_accuracy']
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation acc')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
model.load_weights('/content/drive/Shared drives/personal/weights/resnet_model.31-0.5379.hdf5')
model.evaluate(x_test, y_test)
277/277 [==============================] - 2s 6ms/step
[1.1417856199216325, 0.47653430700302124]
c. Inception V3
def build_inception():
# keras 기본 모델 이용
base_model = inception_v3.InceptionV3(weights=None, include_top=False, input_shape=(256, 256 ,3))
output_model = base_model.output
flatten_model = layers.Flatten()(output_model)
dense_model = layers.Dense(512, activation='relu')(flatten_model)
final_model = layers.Dense(3, activation='softmax')(dense_model)
model = models.Model(inputs=base_model.input, outputs=final_model)
model.compile(optimizer=optimizers.RMSprop(lr=0.0001),
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
model = build_inception()
model.name = 'inception_model'
model.summary()
Model: "inception_model"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_12 (InputLayer) (None, 256, 256, 3) 0
__________________________________________________________________________________________________
conv2d_53 (Conv2D) (None, 127, 127, 32) 864 input_12[0][0]
__________________________________________________________________________________________________
batch_normalization_1 (BatchNor (None, 127, 127, 32) 96 conv2d_53[0][0]
__________________________________________________________________________________________________
activation_295 (Activation) (None, 127, 127, 32) 0 batch_normalization_1[0][0]
__________________________________________________________________________________________________
conv2d_54 (Conv2D) (None, 125, 125, 32) 9216 activation_295[0][0]
__________________________________________________________________________________________________
batch_normalization_2 (BatchNor (None, 125, 125, 32) 96 conv2d_54[0][0]
__________________________________________________________________________________________________
activation_296 (Activation) (None, 125, 125, 32) 0 batch_normalization_2[0][0]
__________________________________________________________________________________________________
conv2d_55 (Conv2D) (None, 125, 125, 64) 18432 activation_296[0][0]
__________________________________________________________________________________________________
batch_normalization_3 (BatchNor (None, 125, 125, 64) 192 conv2d_55[0][0]
__________________________________________________________________________________________________
activation_297 (Activation) (None, 125, 125, 64) 0 batch_normalization_3[0][0]
__________________________________________________________________________________________________
max_pooling2d_18 (MaxPooling2D) (None, 62, 62, 64) 0 activation_297[0][0]
__________________________________________________________________________________________________
conv2d_56 (Conv2D) (None, 62, 62, 80) 5120 max_pooling2d_18[0][0]
__________________________________________________________________________________________________
batch_normalization_4 (BatchNor (None, 62, 62, 80) 240 conv2d_56[0][0]
__________________________________________________________________________________________________
activation_298 (Activation) (None, 62, 62, 80) 0 batch_normalization_4[0][0]
__________________________________________________________________________________________________
conv2d_57 (Conv2D) (None, 60, 60, 192) 138240 activation_298[0][0]
__________________________________________________________________________________________________
batch_normalization_5 (BatchNor (None, 60, 60, 192) 576 conv2d_57[0][0]
__________________________________________________________________________________________________
activation_299 (Activation) (None, 60, 60, 192) 0 batch_normalization_5[0][0]
__________________________________________________________________________________________________
max_pooling2d_19 (MaxPooling2D) (None, 29, 29, 192) 0 activation_299[0][0]
__________________________________________________________________________________________________
conv2d_61 (Conv2D) (None, 29, 29, 64) 12288 max_pooling2d_19[0][0]
__________________________________________________________________________________________________
batch_normalization_9 (BatchNor (None, 29, 29, 64) 192 conv2d_61[0][0]
__________________________________________________________________________________________________
activation_303 (Activation) (None, 29, 29, 64) 0 batch_normalization_9[0][0]
__________________________________________________________________________________________________
conv2d_59 (Conv2D) (None, 29, 29, 48) 9216 max_pooling2d_19[0][0]
__________________________________________________________________________________________________
conv2d_62 (Conv2D) (None, 29, 29, 96) 55296 activation_303[0][0]
__________________________________________________________________________________________________
batch_normalization_7 (BatchNor (None, 29, 29, 48) 144 conv2d_59[0][0]
__________________________________________________________________________________________________
batch_normalization_10 (BatchNo (None, 29, 29, 96) 288 conv2d_62[0][0]
__________________________________________________________________________________________________
activation_301 (Activation) (None, 29, 29, 48) 0 batch_normalization_7[0][0]
__________________________________________________________________________________________________
activation_304 (Activation) (None, 29, 29, 96) 0 batch_normalization_10[0][0]
__________________________________________________________________________________________________
average_pooling2d_1 (AveragePoo (None, 29, 29, 192) 0 max_pooling2d_19[0][0]
__________________________________________________________________________________________________
conv2d_58 (Conv2D) (None, 29, 29, 64) 12288 max_pooling2d_19[0][0]
__________________________________________________________________________________________________
conv2d_60 (Conv2D) (None, 29, 29, 64) 76800 activation_301[0][0]
__________________________________________________________________________________________________
conv2d_63 (Conv2D) (None, 29, 29, 96) 82944 activation_304[0][0]
__________________________________________________________________________________________________
conv2d_64 (Conv2D) (None, 29, 29, 32) 6144 average_pooling2d_1[0][0]
__________________________________________________________________________________________________
batch_normalization_6 (BatchNor (None, 29, 29, 64) 192 conv2d_58[0][0]
__________________________________________________________________________________________________
batch_normalization_8 (BatchNor (None, 29, 29, 64) 192 conv2d_60[0][0]
__________________________________________________________________________________________________
batch_normalization_11 (BatchNo (None, 29, 29, 96) 288 conv2d_63[0][0]
__________________________________________________________________________________________________
batch_normalization_12 (BatchNo (None, 29, 29, 32) 96 conv2d_64[0][0]
__________________________________________________________________________________________________
activation_300 (Activation) (None, 29, 29, 64) 0 batch_normalization_6[0][0]
__________________________________________________________________________________________________
activation_302 (Activation) (None, 29, 29, 64) 0 batch_normalization_8[0][0]
__________________________________________________________________________________________________
activation_305 (Activation) (None, 29, 29, 96) 0 batch_normalization_11[0][0]
__________________________________________________________________________________________________
activation_306 (Activation) (None, 29, 29, 32) 0 batch_normalization_12[0][0]
__________________________________________________________________________________________________
mixed0 (Concatenate) (None, 29, 29, 256) 0 activation_300[0][0]
activation_302[0][0]
activation_305[0][0]
activation_306[0][0]
__________________________________________________________________________________________________
conv2d_68 (Conv2D) (None, 29, 29, 64) 16384 mixed0[0][0]
__________________________________________________________________________________________________
batch_normalization_16 (BatchNo (None, 29, 29, 64) 192 conv2d_68[0][0]
__________________________________________________________________________________________________
activation_310 (Activation) (None, 29, 29, 64) 0 batch_normalization_16[0][0]
__________________________________________________________________________________________________
conv2d_66 (Conv2D) (None, 29, 29, 48) 12288 mixed0[0][0]
__________________________________________________________________________________________________
conv2d_69 (Conv2D) (None, 29, 29, 96) 55296 activation_310[0][0]
__________________________________________________________________________________________________
batch_normalization_14 (BatchNo (None, 29, 29, 48) 144 conv2d_66[0][0]
__________________________________________________________________________________________________
batch_normalization_17 (BatchNo (None, 29, 29, 96) 288 conv2d_69[0][0]
__________________________________________________________________________________________________
activation_308 (Activation) (None, 29, 29, 48) 0 batch_normalization_14[0][0]
__________________________________________________________________________________________________
activation_311 (Activation) (None, 29, 29, 96) 0 batch_normalization_17[0][0]
__________________________________________________________________________________________________
average_pooling2d_2 (AveragePoo (None, 29, 29, 256) 0 mixed0[0][0]
__________________________________________________________________________________________________
conv2d_65 (Conv2D) (None, 29, 29, 64) 16384 mixed0[0][0]
__________________________________________________________________________________________________
conv2d_67 (Conv2D) (None, 29, 29, 64) 76800 activation_308[0][0]
__________________________________________________________________________________________________
conv2d_70 (Conv2D) (None, 29, 29, 96) 82944 activation_311[0][0]
__________________________________________________________________________________________________
conv2d_71 (Conv2D) (None, 29, 29, 64) 16384 average_pooling2d_2[0][0]
__________________________________________________________________________________________________
batch_normalization_13 (BatchNo (None, 29, 29, 64) 192 conv2d_65[0][0]
__________________________________________________________________________________________________
batch_normalization_15 (BatchNo (None, 29, 29, 64) 192 conv2d_67[0][0]
__________________________________________________________________________________________________
batch_normalization_18 (BatchNo (None, 29, 29, 96) 288 conv2d_70[0][0]
__________________________________________________________________________________________________
batch_normalization_19 (BatchNo (None, 29, 29, 64) 192 conv2d_71[0][0]
__________________________________________________________________________________________________
activation_307 (Activation) (None, 29, 29, 64) 0 batch_normalization_13[0][0]
__________________________________________________________________________________________________
activation_309 (Activation) (None, 29, 29, 64) 0 batch_normalization_15[0][0]
__________________________________________________________________________________________________
activation_312 (Activation) (None, 29, 29, 96) 0 batch_normalization_18[0][0]
__________________________________________________________________________________________________
activation_313 (Activation) (None, 29, 29, 64) 0 batch_normalization_19[0][0]
__________________________________________________________________________________________________
mixed1 (Concatenate) (None, 29, 29, 288) 0 activation_307[0][0]
activation_309[0][0]
activation_312[0][0]
activation_313[0][0]
__________________________________________________________________________________________________
conv2d_75 (Conv2D) (None, 29, 29, 64) 18432 mixed1[0][0]
__________________________________________________________________________________________________
batch_normalization_23 (BatchNo (None, 29, 29, 64) 192 conv2d_75[0][0]
__________________________________________________________________________________________________
activation_317 (Activation) (None, 29, 29, 64) 0 batch_normalization_23[0][0]
__________________________________________________________________________________________________
conv2d_73 (Conv2D) (None, 29, 29, 48) 13824 mixed1[0][0]
__________________________________________________________________________________________________
conv2d_76 (Conv2D) (None, 29, 29, 96) 55296 activation_317[0][0]
__________________________________________________________________________________________________
batch_normalization_21 (BatchNo (None, 29, 29, 48) 144 conv2d_73[0][0]
__________________________________________________________________________________________________
batch_normalization_24 (BatchNo (None, 29, 29, 96) 288 conv2d_76[0][0]
__________________________________________________________________________________________________
activation_315 (Activation) (None, 29, 29, 48) 0 batch_normalization_21[0][0]
__________________________________________________________________________________________________
activation_318 (Activation) (None, 29, 29, 96) 0 batch_normalization_24[0][0]
__________________________________________________________________________________________________
average_pooling2d_3 (AveragePoo (None, 29, 29, 288) 0 mixed1[0][0]
__________________________________________________________________________________________________
conv2d_72 (Conv2D) (None, 29, 29, 64) 18432 mixed1[0][0]
__________________________________________________________________________________________________
conv2d_74 (Conv2D) (None, 29, 29, 64) 76800 activation_315[0][0]
__________________________________________________________________________________________________
conv2d_77 (Conv2D) (None, 29, 29, 96) 82944 activation_318[0][0]
__________________________________________________________________________________________________
conv2d_78 (Conv2D) (None, 29, 29, 64) 18432 average_pooling2d_3[0][0]
__________________________________________________________________________________________________
batch_normalization_20 (BatchNo (None, 29, 29, 64) 192 conv2d_72[0][0]
__________________________________________________________________________________________________
batch_normalization_22 (BatchNo (None, 29, 29, 64) 192 conv2d_74[0][0]
__________________________________________________________________________________________________
batch_normalization_25 (BatchNo (None, 29, 29, 96) 288 conv2d_77[0][0]
__________________________________________________________________________________________________
batch_normalization_26 (BatchNo (None, 29, 29, 64) 192 conv2d_78[0][0]
__________________________________________________________________________________________________
activation_314 (Activation) (None, 29, 29, 64) 0 batch_normalization_20[0][0]
__________________________________________________________________________________________________
activation_316 (Activation) (None, 29, 29, 64) 0 batch_normalization_22[0][0]
__________________________________________________________________________________________________
activation_319 (Activation) (None, 29, 29, 96) 0 batch_normalization_25[0][0]
__________________________________________________________________________________________________
activation_320 (Activation) (None, 29, 29, 64) 0 batch_normalization_26[0][0]
__________________________________________________________________________________________________
mixed2 (Concatenate) (None, 29, 29, 288) 0 activation_314[0][0]
activation_316[0][0]
activation_319[0][0]
activation_320[0][0]
__________________________________________________________________________________________________
conv2d_80 (Conv2D) (None, 29, 29, 64) 18432 mixed2[0][0]
__________________________________________________________________________________________________
batch_normalization_28 (BatchNo (None, 29, 29, 64) 192 conv2d_80[0][0]
__________________________________________________________________________________________________
activation_322 (Activation) (None, 29, 29, 64) 0 batch_normalization_28[0][0]
__________________________________________________________________________________________________
conv2d_81 (Conv2D) (None, 29, 29, 96) 55296 activation_322[0][0]
__________________________________________________________________________________________________
batch_normalization_29 (BatchNo (None, 29, 29, 96) 288 conv2d_81[0][0]
__________________________________________________________________________________________________
activation_323 (Activation) (None, 29, 29, 96) 0 batch_normalization_29[0][0]
__________________________________________________________________________________________________
conv2d_79 (Conv2D) (None, 14, 14, 384) 995328 mixed2[0][0]
__________________________________________________________________________________________________
conv2d_82 (Conv2D) (None, 14, 14, 96) 82944 activation_323[0][0]
__________________________________________________________________________________________________
batch_normalization_27 (BatchNo (None, 14, 14, 384) 1152 conv2d_79[0][0]
__________________________________________________________________________________________________
batch_normalization_30 (BatchNo (None, 14, 14, 96) 288 conv2d_82[0][0]
__________________________________________________________________________________________________
activation_321 (Activation) (None, 14, 14, 384) 0 batch_normalization_27[0][0]
__________________________________________________________________________________________________
activation_324 (Activation) (None, 14, 14, 96) 0 batch_normalization_30[0][0]
__________________________________________________________________________________________________
max_pooling2d_20 (MaxPooling2D) (None, 14, 14, 288) 0 mixed2[0][0]
__________________________________________________________________________________________________
mixed3 (Concatenate) (None, 14, 14, 768) 0 activation_321[0][0]
activation_324[0][0]
max_pooling2d_20[0][0]
__________________________________________________________________________________________________
conv2d_87 (Conv2D) (None, 14, 14, 128) 98304 mixed3[0][0]
__________________________________________________________________________________________________
batch_normalization_35 (BatchNo (None, 14, 14, 128) 384 conv2d_87[0][0]
__________________________________________________________________________________________________
activation_329 (Activation) (None, 14, 14, 128) 0 batch_normalization_35[0][0]
__________________________________________________________________________________________________
conv2d_88 (Conv2D) (None, 14, 14, 128) 114688 activation_329[0][0]
__________________________________________________________________________________________________
batch_normalization_36 (BatchNo (None, 14, 14, 128) 384 conv2d_88[0][0]
__________________________________________________________________________________________________
activation_330 (Activation) (None, 14, 14, 128) 0 batch_normalization_36[0][0]
__________________________________________________________________________________________________
conv2d_84 (Conv2D) (None, 14, 14, 128) 98304 mixed3[0][0]
__________________________________________________________________________________________________
conv2d_89 (Conv2D) (None, 14, 14, 128) 114688 activation_330[0][0]
__________________________________________________________________________________________________
batch_normalization_32 (BatchNo (None, 14, 14, 128) 384 conv2d_84[0][0]
__________________________________________________________________________________________________
batch_normalization_37 (BatchNo (None, 14, 14, 128) 384 conv2d_89[0][0]
__________________________________________________________________________________________________
activation_326 (Activation) (None, 14, 14, 128) 0 batch_normalization_32[0][0]
__________________________________________________________________________________________________
activation_331 (Activation) (None, 14, 14, 128) 0 batch_normalization_37[0][0]
__________________________________________________________________________________________________
conv2d_85 (Conv2D) (None, 14, 14, 128) 114688 activation_326[0][0]
__________________________________________________________________________________________________
conv2d_90 (Conv2D) (None, 14, 14, 128) 114688 activation_331[0][0]
__________________________________________________________________________________________________
batch_normalization_33 (BatchNo (None, 14, 14, 128) 384 conv2d_85[0][0]
__________________________________________________________________________________________________
batch_normalization_38 (BatchNo (None, 14, 14, 128) 384 conv2d_90[0][0]
__________________________________________________________________________________________________
activation_327 (Activation) (None, 14, 14, 128) 0 batch_normalization_33[0][0]
__________________________________________________________________________________________________
activation_332 (Activation) (None, 14, 14, 128) 0 batch_normalization_38[0][0]
__________________________________________________________________________________________________
average_pooling2d_4 (AveragePoo (None, 14, 14, 768) 0 mixed3[0][0]
__________________________________________________________________________________________________
conv2d_83 (Conv2D) (None, 14, 14, 192) 147456 mixed3[0][0]
__________________________________________________________________________________________________
conv2d_86 (Conv2D) (None, 14, 14, 192) 172032 activation_327[0][0]
__________________________________________________________________________________________________
conv2d_91 (Conv2D) (None, 14, 14, 192) 172032 activation_332[0][0]
__________________________________________________________________________________________________
conv2d_92 (Conv2D) (None, 14, 14, 192) 147456 average_pooling2d_4[0][0]
__________________________________________________________________________________________________
batch_normalization_31 (BatchNo (None, 14, 14, 192) 576 conv2d_83[0][0]
__________________________________________________________________________________________________
batch_normalization_34 (BatchNo (None, 14, 14, 192) 576 conv2d_86[0][0]
__________________________________________________________________________________________________
batch_normalization_39 (BatchNo (None, 14, 14, 192) 576 conv2d_91[0][0]
__________________________________________________________________________________________________
batch_normalization_40 (BatchNo (None, 14, 14, 192) 576 conv2d_92[0][0]
__________________________________________________________________________________________________
activation_325 (Activation) (None, 14, 14, 192) 0 batch_normalization_31[0][0]
__________________________________________________________________________________________________
activation_328 (Activation) (None, 14, 14, 192) 0 batch_normalization_34[0][0]
__________________________________________________________________________________________________
activation_333 (Activation) (None, 14, 14, 192) 0 batch_normalization_39[0][0]
__________________________________________________________________________________________________
activation_334 (Activation) (None, 14, 14, 192) 0 batch_normalization_40[0][0]
__________________________________________________________________________________________________
mixed4 (Concatenate) (None, 14, 14, 768) 0 activation_325[0][0]
activation_328[0][0]
activation_333[0][0]
activation_334[0][0]
__________________________________________________________________________________________________
conv2d_97 (Conv2D) (None, 14, 14, 160) 122880 mixed4[0][0]
__________________________________________________________________________________________________
batch_normalization_45 (BatchNo (None, 14, 14, 160) 480 conv2d_97[0][0]
__________________________________________________________________________________________________
activation_339 (Activation) (None, 14, 14, 160) 0 batch_normalization_45[0][0]
__________________________________________________________________________________________________
conv2d_98 (Conv2D) (None, 14, 14, 160) 179200 activation_339[0][0]
__________________________________________________________________________________________________
batch_normalization_46 (BatchNo (None, 14, 14, 160) 480 conv2d_98[0][0]
__________________________________________________________________________________________________
activation_340 (Activation) (None, 14, 14, 160) 0 batch_normalization_46[0][0]
__________________________________________________________________________________________________
conv2d_94 (Conv2D) (None, 14, 14, 160) 122880 mixed4[0][0]
__________________________________________________________________________________________________
conv2d_99 (Conv2D) (None, 14, 14, 160) 179200 activation_340[0][0]
__________________________________________________________________________________________________
batch_normalization_42 (BatchNo (None, 14, 14, 160) 480 conv2d_94[0][0]
__________________________________________________________________________________________________
batch_normalization_47 (BatchNo (None, 14, 14, 160) 480 conv2d_99[0][0]
__________________________________________________________________________________________________
activation_336 (Activation) (None, 14, 14, 160) 0 batch_normalization_42[0][0]
__________________________________________________________________________________________________
activation_341 (Activation) (None, 14, 14, 160) 0 batch_normalization_47[0][0]
__________________________________________________________________________________________________
conv2d_95 (Conv2D) (None, 14, 14, 160) 179200 activation_336[0][0]
__________________________________________________________________________________________________
conv2d_100 (Conv2D) (None, 14, 14, 160) 179200 activation_341[0][0]
__________________________________________________________________________________________________
batch_normalization_43 (BatchNo (None, 14, 14, 160) 480 conv2d_95[0][0]
__________________________________________________________________________________________________
batch_normalization_48 (BatchNo (None, 14, 14, 160) 480 conv2d_100[0][0]
__________________________________________________________________________________________________
activation_337 (Activation) (None, 14, 14, 160) 0 batch_normalization_43[0][0]
__________________________________________________________________________________________________
activation_342 (Activation) (None, 14, 14, 160) 0 batch_normalization_48[0][0]
__________________________________________________________________________________________________
average_pooling2d_5 (AveragePoo (None, 14, 14, 768) 0 mixed4[0][0]
__________________________________________________________________________________________________
conv2d_93 (Conv2D) (None, 14, 14, 192) 147456 mixed4[0][0]
__________________________________________________________________________________________________
conv2d_96 (Conv2D) (None, 14, 14, 192) 215040 activation_337[0][0]
__________________________________________________________________________________________________
conv2d_101 (Conv2D) (None, 14, 14, 192) 215040 activation_342[0][0]
__________________________________________________________________________________________________
conv2d_102 (Conv2D) (None, 14, 14, 192) 147456 average_pooling2d_5[0][0]
__________________________________________________________________________________________________
batch_normalization_41 (BatchNo (None, 14, 14, 192) 576 conv2d_93[0][0]
__________________________________________________________________________________________________
batch_normalization_44 (BatchNo (None, 14, 14, 192) 576 conv2d_96[0][0]
__________________________________________________________________________________________________
batch_normalization_49 (BatchNo (None, 14, 14, 192) 576 conv2d_101[0][0]
__________________________________________________________________________________________________
batch_normalization_50 (BatchNo (None, 14, 14, 192) 576 conv2d_102[0][0]
__________________________________________________________________________________________________
activation_335 (Activation) (None, 14, 14, 192) 0 batch_normalization_41[0][0]
__________________________________________________________________________________________________
activation_338 (Activation) (None, 14, 14, 192) 0 batch_normalization_44[0][0]
__________________________________________________________________________________________________
activation_343 (Activation) (None, 14, 14, 192) 0 batch_normalization_49[0][0]
__________________________________________________________________________________________________
activation_344 (Activation) (None, 14, 14, 192) 0 batch_normalization_50[0][0]
__________________________________________________________________________________________________
mixed5 (Concatenate) (None, 14, 14, 768) 0 activation_335[0][0]
activation_338[0][0]
activation_343[0][0]
activation_344[0][0]
__________________________________________________________________________________________________
conv2d_107 (Conv2D) (None, 14, 14, 160) 122880 mixed5[0][0]
__________________________________________________________________________________________________
batch_normalization_55 (BatchNo (None, 14, 14, 160) 480 conv2d_107[0][0]
__________________________________________________________________________________________________
activation_349 (Activation) (None, 14, 14, 160) 0 batch_normalization_55[0][0]
__________________________________________________________________________________________________
conv2d_108 (Conv2D) (None, 14, 14, 160) 179200 activation_349[0][0]
__________________________________________________________________________________________________
batch_normalization_56 (BatchNo (None, 14, 14, 160) 480 conv2d_108[0][0]
__________________________________________________________________________________________________
activation_350 (Activation) (None, 14, 14, 160) 0 batch_normalization_56[0][0]
__________________________________________________________________________________________________
conv2d_104 (Conv2D) (None, 14, 14, 160) 122880 mixed5[0][0]
__________________________________________________________________________________________________
conv2d_109 (Conv2D) (None, 14, 14, 160) 179200 activation_350[0][0]
__________________________________________________________________________________________________
batch_normalization_52 (BatchNo (None, 14, 14, 160) 480 conv2d_104[0][0]
__________________________________________________________________________________________________
batch_normalization_57 (BatchNo (None, 14, 14, 160) 480 conv2d_109[0][0]
__________________________________________________________________________________________________
activation_346 (Activation) (None, 14, 14, 160) 0 batch_normalization_52[0][0]
__________________________________________________________________________________________________
activation_351 (Activation) (None, 14, 14, 160) 0 batch_normalization_57[0][0]
__________________________________________________________________________________________________
conv2d_105 (Conv2D) (None, 14, 14, 160) 179200 activation_346[0][0]
__________________________________________________________________________________________________
conv2d_110 (Conv2D) (None, 14, 14, 160) 179200 activation_351[0][0]
__________________________________________________________________________________________________
batch_normalization_53 (BatchNo (None, 14, 14, 160) 480 conv2d_105[0][0]
__________________________________________________________________________________________________
batch_normalization_58 (BatchNo (None, 14, 14, 160) 480 conv2d_110[0][0]
__________________________________________________________________________________________________
activation_347 (Activation) (None, 14, 14, 160) 0 batch_normalization_53[0][0]
__________________________________________________________________________________________________
activation_352 (Activation) (None, 14, 14, 160) 0 batch_normalization_58[0][0]
__________________________________________________________________________________________________
average_pooling2d_6 (AveragePoo (None, 14, 14, 768) 0 mixed5[0][0]
__________________________________________________________________________________________________
conv2d_103 (Conv2D) (None, 14, 14, 192) 147456 mixed5[0][0]
__________________________________________________________________________________________________
conv2d_106 (Conv2D) (None, 14, 14, 192) 215040 activation_347[0][0]
__________________________________________________________________________________________________
conv2d_111 (Conv2D) (None, 14, 14, 192) 215040 activation_352[0][0]
__________________________________________________________________________________________________
conv2d_112 (Conv2D) (None, 14, 14, 192) 147456 average_pooling2d_6[0][0]
__________________________________________________________________________________________________
batch_normalization_51 (BatchNo (None, 14, 14, 192) 576 conv2d_103[0][0]
__________________________________________________________________________________________________
batch_normalization_54 (BatchNo (None, 14, 14, 192) 576 conv2d_106[0][0]
__________________________________________________________________________________________________
batch_normalization_59 (BatchNo (None, 14, 14, 192) 576 conv2d_111[0][0]
__________________________________________________________________________________________________
batch_normalization_60 (BatchNo (None, 14, 14, 192) 576 conv2d_112[0][0]
__________________________________________________________________________________________________
activation_345 (Activation) (None, 14, 14, 192) 0 batch_normalization_51[0][0]
__________________________________________________________________________________________________
activation_348 (Activation) (None, 14, 14, 192) 0 batch_normalization_54[0][0]
__________________________________________________________________________________________________
activation_353 (Activation) (None, 14, 14, 192) 0 batch_normalization_59[0][0]
__________________________________________________________________________________________________
activation_354 (Activation) (None, 14, 14, 192) 0 batch_normalization_60[0][0]
__________________________________________________________________________________________________
mixed6 (Concatenate) (None, 14, 14, 768) 0 activation_345[0][0]
activation_348[0][0]
activation_353[0][0]
activation_354[0][0]
__________________________________________________________________________________________________
conv2d_117 (Conv2D) (None, 14, 14, 192) 147456 mixed6[0][0]
__________________________________________________________________________________________________
batch_normalization_65 (BatchNo (None, 14, 14, 192) 576 conv2d_117[0][0]
__________________________________________________________________________________________________
activation_359 (Activation) (None, 14, 14, 192) 0 batch_normalization_65[0][0]
__________________________________________________________________________________________________
conv2d_118 (Conv2D) (None, 14, 14, 192) 258048 activation_359[0][0]
__________________________________________________________________________________________________
batch_normalization_66 (BatchNo (None, 14, 14, 192) 576 conv2d_118[0][0]
__________________________________________________________________________________________________
activation_360 (Activation) (None, 14, 14, 192) 0 batch_normalization_66[0][0]
__________________________________________________________________________________________________
conv2d_114 (Conv2D) (None, 14, 14, 192) 147456 mixed6[0][0]
__________________________________________________________________________________________________
conv2d_119 (Conv2D) (None, 14, 14, 192) 258048 activation_360[0][0]
__________________________________________________________________________________________________
batch_normalization_62 (BatchNo (None, 14, 14, 192) 576 conv2d_114[0][0]
__________________________________________________________________________________________________
batch_normalization_67 (BatchNo (None, 14, 14, 192) 576 conv2d_119[0][0]
__________________________________________________________________________________________________
activation_356 (Activation) (None, 14, 14, 192) 0 batch_normalization_62[0][0]
__________________________________________________________________________________________________
activation_361 (Activation) (None, 14, 14, 192) 0 batch_normalization_67[0][0]
__________________________________________________________________________________________________
conv2d_115 (Conv2D) (None, 14, 14, 192) 258048 activation_356[0][0]
__________________________________________________________________________________________________
conv2d_120 (Conv2D) (None, 14, 14, 192) 258048 activation_361[0][0]
__________________________________________________________________________________________________
batch_normalization_63 (BatchNo (None, 14, 14, 192) 576 conv2d_115[0][0]
__________________________________________________________________________________________________
batch_normalization_68 (BatchNo (None, 14, 14, 192) 576 conv2d_120[0][0]
__________________________________________________________________________________________________
activation_357 (Activation) (None, 14, 14, 192) 0 batch_normalization_63[0][0]
__________________________________________________________________________________________________
activation_362 (Activation) (None, 14, 14, 192) 0 batch_normalization_68[0][0]
__________________________________________________________________________________________________
average_pooling2d_7 (AveragePoo (None, 14, 14, 768) 0 mixed6[0][0]
__________________________________________________________________________________________________
conv2d_113 (Conv2D) (None, 14, 14, 192) 147456 mixed6[0][0]
__________________________________________________________________________________________________
conv2d_116 (Conv2D) (None, 14, 14, 192) 258048 activation_357[0][0]
__________________________________________________________________________________________________
conv2d_121 (Conv2D) (None, 14, 14, 192) 258048 activation_362[0][0]
__________________________________________________________________________________________________
conv2d_122 (Conv2D) (None, 14, 14, 192) 147456 average_pooling2d_7[0][0]
__________________________________________________________________________________________________
batch_normalization_61 (BatchNo (None, 14, 14, 192) 576 conv2d_113[0][0]
__________________________________________________________________________________________________
batch_normalization_64 (BatchNo (None, 14, 14, 192) 576 conv2d_116[0][0]
__________________________________________________________________________________________________
batch_normalization_69 (BatchNo (None, 14, 14, 192) 576 conv2d_121[0][0]
__________________________________________________________________________________________________
batch_normalization_70 (BatchNo (None, 14, 14, 192) 576 conv2d_122[0][0]
__________________________________________________________________________________________________
activation_355 (Activation) (None, 14, 14, 192) 0 batch_normalization_61[0][0]
__________________________________________________________________________________________________
activation_358 (Activation) (None, 14, 14, 192) 0 batch_normalization_64[0][0]
__________________________________________________________________________________________________
activation_363 (Activation) (None, 14, 14, 192) 0 batch_normalization_69[0][0]
__________________________________________________________________________________________________
activation_364 (Activation) (None, 14, 14, 192) 0 batch_normalization_70[0][0]
__________________________________________________________________________________________________
mixed7 (Concatenate) (None, 14, 14, 768) 0 activation_355[0][0]
activation_358[0][0]
activation_363[0][0]
activation_364[0][0]
__________________________________________________________________________________________________
conv2d_125 (Conv2D) (None, 14, 14, 192) 147456 mixed7[0][0]
__________________________________________________________________________________________________
batch_normalization_73 (BatchNo (None, 14, 14, 192) 576 conv2d_125[0][0]
__________________________________________________________________________________________________
activation_367 (Activation) (None, 14, 14, 192) 0 batch_normalization_73[0][0]
__________________________________________________________________________________________________
conv2d_126 (Conv2D) (None, 14, 14, 192) 258048 activation_367[0][0]
__________________________________________________________________________________________________
batch_normalization_74 (BatchNo (None, 14, 14, 192) 576 conv2d_126[0][0]
__________________________________________________________________________________________________
activation_368 (Activation) (None, 14, 14, 192) 0 batch_normalization_74[0][0]
__________________________________________________________________________________________________
conv2d_123 (Conv2D) (None, 14, 14, 192) 147456 mixed7[0][0]
__________________________________________________________________________________________________
conv2d_127 (Conv2D) (None, 14, 14, 192) 258048 activation_368[0][0]
__________________________________________________________________________________________________
batch_normalization_71 (BatchNo (None, 14, 14, 192) 576 conv2d_123[0][0]
__________________________________________________________________________________________________
batch_normalization_75 (BatchNo (None, 14, 14, 192) 576 conv2d_127[0][0]
__________________________________________________________________________________________________
activation_365 (Activation) (None, 14, 14, 192) 0 batch_normalization_71[0][0]
__________________________________________________________________________________________________
activation_369 (Activation) (None, 14, 14, 192) 0 batch_normalization_75[0][0]
__________________________________________________________________________________________________
conv2d_124 (Conv2D) (None, 6, 6, 320) 552960 activation_365[0][0]
__________________________________________________________________________________________________
conv2d_128 (Conv2D) (None, 6, 6, 192) 331776 activation_369[0][0]
__________________________________________________________________________________________________
batch_normalization_72 (BatchNo (None, 6, 6, 320) 960 conv2d_124[0][0]
__________________________________________________________________________________________________
batch_normalization_76 (BatchNo (None, 6, 6, 192) 576 conv2d_128[0][0]
__________________________________________________________________________________________________
activation_366 (Activation) (None, 6, 6, 320) 0 batch_normalization_72[0][0]
__________________________________________________________________________________________________
activation_370 (Activation) (None, 6, 6, 192) 0 batch_normalization_76[0][0]
__________________________________________________________________________________________________
max_pooling2d_21 (MaxPooling2D) (None, 6, 6, 768) 0 mixed7[0][0]
__________________________________________________________________________________________________
mixed8 (Concatenate) (None, 6, 6, 1280) 0 activation_366[0][0]
activation_370[0][0]
max_pooling2d_21[0][0]
__________________________________________________________________________________________________
conv2d_133 (Conv2D) (None, 6, 6, 448) 573440 mixed8[0][0]
__________________________________________________________________________________________________
batch_normalization_81 (BatchNo (None, 6, 6, 448) 1344 conv2d_133[0][0]
__________________________________________________________________________________________________
activation_375 (Activation) (None, 6, 6, 448) 0 batch_normalization_81[0][0]
__________________________________________________________________________________________________
conv2d_130 (Conv2D) (None, 6, 6, 384) 491520 mixed8[0][0]
__________________________________________________________________________________________________
conv2d_134 (Conv2D) (None, 6, 6, 384) 1548288 activation_375[0][0]
__________________________________________________________________________________________________
batch_normalization_78 (BatchNo (None, 6, 6, 384) 1152 conv2d_130[0][0]
__________________________________________________________________________________________________
batch_normalization_82 (BatchNo (None, 6, 6, 384) 1152 conv2d_134[0][0]
__________________________________________________________________________________________________
activation_372 (Activation) (None, 6, 6, 384) 0 batch_normalization_78[0][0]
__________________________________________________________________________________________________
activation_376 (Activation) (None, 6, 6, 384) 0 batch_normalization_82[0][0]
__________________________________________________________________________________________________
conv2d_131 (Conv2D) (None, 6, 6, 384) 442368 activation_372[0][0]
__________________________________________________________________________________________________
conv2d_132 (Conv2D) (None, 6, 6, 384) 442368 activation_372[0][0]
__________________________________________________________________________________________________
conv2d_135 (Conv2D) (None, 6, 6, 384) 442368 activation_376[0][0]
__________________________________________________________________________________________________
conv2d_136 (Conv2D) (None, 6, 6, 384) 442368 activation_376[0][0]
__________________________________________________________________________________________________
average_pooling2d_8 (AveragePoo (None, 6, 6, 1280) 0 mixed8[0][0]
__________________________________________________________________________________________________
conv2d_129 (Conv2D) (None, 6, 6, 320) 409600 mixed8[0][0]
__________________________________________________________________________________________________
batch_normalization_79 (BatchNo (None, 6, 6, 384) 1152 conv2d_131[0][0]
__________________________________________________________________________________________________
batch_normalization_80 (BatchNo (None, 6, 6, 384) 1152 conv2d_132[0][0]
__________________________________________________________________________________________________
batch_normalization_83 (BatchNo (None, 6, 6, 384) 1152 conv2d_135[0][0]
__________________________________________________________________________________________________
batch_normalization_84 (BatchNo (None, 6, 6, 384) 1152 conv2d_136[0][0]
__________________________________________________________________________________________________
conv2d_137 (Conv2D) (None, 6, 6, 192) 245760 average_pooling2d_8[0][0]
__________________________________________________________________________________________________
batch_normalization_77 (BatchNo (None, 6, 6, 320) 960 conv2d_129[0][0]
__________________________________________________________________________________________________
activation_373 (Activation) (None, 6, 6, 384) 0 batch_normalization_79[0][0]
__________________________________________________________________________________________________
activation_374 (Activation) (None, 6, 6, 384) 0 batch_normalization_80[0][0]
__________________________________________________________________________________________________
activation_377 (Activation) (None, 6, 6, 384) 0 batch_normalization_83[0][0]
__________________________________________________________________________________________________
activation_378 (Activation) (None, 6, 6, 384) 0 batch_normalization_84[0][0]
__________________________________________________________________________________________________
batch_normalization_85 (BatchNo (None, 6, 6, 192) 576 conv2d_137[0][0]
__________________________________________________________________________________________________
activation_371 (Activation) (None, 6, 6, 320) 0 batch_normalization_77[0][0]
__________________________________________________________________________________________________
mixed9_0 (Concatenate) (None, 6, 6, 768) 0 activation_373[0][0]
activation_374[0][0]
__________________________________________________________________________________________________
concatenate_9 (Concatenate) (None, 6, 6, 768) 0 activation_377[0][0]
activation_378[0][0]
__________________________________________________________________________________________________
activation_379 (Activation) (None, 6, 6, 192) 0 batch_normalization_85[0][0]
__________________________________________________________________________________________________
mixed9 (Concatenate) (None, 6, 6, 2048) 0 activation_371[0][0]
mixed9_0[0][0]
concatenate_9[0][0]
activation_379[0][0]
__________________________________________________________________________________________________
conv2d_142 (Conv2D) (None, 6, 6, 448) 917504 mixed9[0][0]
__________________________________________________________________________________________________
batch_normalization_90 (BatchNo (None, 6, 6, 448) 1344 conv2d_142[0][0]
__________________________________________________________________________________________________
activation_384 (Activation) (None, 6, 6, 448) 0 batch_normalization_90[0][0]
__________________________________________________________________________________________________
conv2d_139 (Conv2D) (None, 6, 6, 384) 786432 mixed9[0][0]
__________________________________________________________________________________________________
conv2d_143 (Conv2D) (None, 6, 6, 384) 1548288 activation_384[0][0]
__________________________________________________________________________________________________
batch_normalization_87 (BatchNo (None, 6, 6, 384) 1152 conv2d_139[0][0]
__________________________________________________________________________________________________
batch_normalization_91 (BatchNo (None, 6, 6, 384) 1152 conv2d_143[0][0]
__________________________________________________________________________________________________
activation_381 (Activation) (None, 6, 6, 384) 0 batch_normalization_87[0][0]
__________________________________________________________________________________________________
activation_385 (Activation) (None, 6, 6, 384) 0 batch_normalization_91[0][0]
__________________________________________________________________________________________________
conv2d_140 (Conv2D) (None, 6, 6, 384) 442368 activation_381[0][0]
__________________________________________________________________________________________________
conv2d_141 (Conv2D) (None, 6, 6, 384) 442368 activation_381[0][0]
__________________________________________________________________________________________________
conv2d_144 (Conv2D) (None, 6, 6, 384) 442368 activation_385[0][0]
__________________________________________________________________________________________________
conv2d_145 (Conv2D) (None, 6, 6, 384) 442368 activation_385[0][0]
__________________________________________________________________________________________________
average_pooling2d_9 (AveragePoo (None, 6, 6, 2048) 0 mixed9[0][0]
__________________________________________________________________________________________________
conv2d_138 (Conv2D) (None, 6, 6, 320) 655360 mixed9[0][0]
__________________________________________________________________________________________________
batch_normalization_88 (BatchNo (None, 6, 6, 384) 1152 conv2d_140[0][0]
__________________________________________________________________________________________________
batch_normalization_89 (BatchNo (None, 6, 6, 384) 1152 conv2d_141[0][0]
__________________________________________________________________________________________________
batch_normalization_92 (BatchNo (None, 6, 6, 384) 1152 conv2d_144[0][0]
__________________________________________________________________________________________________
batch_normalization_93 (BatchNo (None, 6, 6, 384) 1152 conv2d_145[0][0]
__________________________________________________________________________________________________
conv2d_146 (Conv2D) (None, 6, 6, 192) 393216 average_pooling2d_9[0][0]
__________________________________________________________________________________________________
batch_normalization_86 (BatchNo (None, 6, 6, 320) 960 conv2d_138[0][0]
__________________________________________________________________________________________________
activation_382 (Activation) (None, 6, 6, 384) 0 batch_normalization_88[0][0]
__________________________________________________________________________________________________
activation_383 (Activation) (None, 6, 6, 384) 0 batch_normalization_89[0][0]
__________________________________________________________________________________________________
activation_386 (Activation) (None, 6, 6, 384) 0 batch_normalization_92[0][0]
__________________________________________________________________________________________________
activation_387 (Activation) (None, 6, 6, 384) 0 batch_normalization_93[0][0]
__________________________________________________________________________________________________
batch_normalization_94 (BatchNo (None, 6, 6, 192) 576 conv2d_146[0][0]
__________________________________________________________________________________________________
activation_380 (Activation) (None, 6, 6, 320) 0 batch_normalization_86[0][0]
__________________________________________________________________________________________________
mixed9_1 (Concatenate) (None, 6, 6, 768) 0 activation_382[0][0]
activation_383[0][0]
__________________________________________________________________________________________________
concatenate_10 (Concatenate) (None, 6, 6, 768) 0 activation_386[0][0]
activation_387[0][0]
__________________________________________________________________________________________________
activation_388 (Activation) (None, 6, 6, 192) 0 batch_normalization_94[0][0]
__________________________________________________________________________________________________
mixed10 (Concatenate) (None, 6, 6, 2048) 0 activation_380[0][0]
mixed9_1[0][0]
concatenate_10[0][0]
activation_388[0][0]
__________________________________________________________________________________________________
flatten_8 (Flatten) (None, 73728) 0 mixed10[0][0]
__________________________________________________________________________________________________
dense_18 (Dense) (None, 512) 37749248 flatten_8[0][0]
__________________________________________________________________________________________________
dense_19 (Dense) (None, 3) 1539 dense_18[0][0]
==================================================================================================
Total params: 59,553,571
Trainable params: 59,519,139
Non-trainable params: 34,432
__________________________________________________________________________________________________
history = model_train(model, 40)
Train on 829 samples, validate on 277 samples
Epoch 1/40
829/829 [==============================] - 30s 36ms/step - loss: 3.7299 - accuracy: 0.3932 - val_loss: 1.0825 - val_accuracy: 0.4910
Epoch 2/40
829/829 [==============================] - 12s 14ms/step - loss: 1.2540 - accuracy: 0.4729 - val_loss: 1.0804 - val_accuracy: 0.4910
Epoch 3/40
829/829 [==============================] - 11s 14ms/step - loss: 1.1731 - accuracy: 0.4524 - val_loss: 1.1040 - val_accuracy: 0.4910
Epoch 4/40
829/829 [==============================] - 11s 14ms/step - loss: 1.0617 - accuracy: 0.4632 - val_loss: 1.0473 - val_accuracy: 0.4910
Epoch 5/40
829/829 [==============================] - 12s 14ms/step - loss: 1.0231 - accuracy: 0.5066 - val_loss: 1.0410 - val_accuracy: 0.4910
Epoch 6/40
829/829 [==============================] - 12s 14ms/step - loss: 0.9809 - accuracy: 0.5271 - val_loss: 1.0515 - val_accuracy: 0.4838
Epoch 7/40
829/829 [==============================] - 12s 14ms/step - loss: 0.9229 - accuracy: 0.5754 - val_loss: 1.0551 - val_accuracy: 0.4260
Epoch 8/40
829/829 [==============================] - 11s 14ms/step - loss: 0.8852 - accuracy: 0.6055 - val_loss: 1.0078 - val_accuracy: 0.4982
Epoch 9/40
829/829 [==============================] - 11s 14ms/step - loss: 0.8556 - accuracy: 0.6019 - val_loss: 1.1044 - val_accuracy: 0.4513
Epoch 10/40
829/829 [==============================] - 11s 14ms/step - loss: 0.8235 - accuracy: 0.6333 - val_loss: 1.2843 - val_accuracy: 0.5235
Epoch 11/40
829/829 [==============================] - 11s 14ms/step - loss: 0.7848 - accuracy: 0.6598 - val_loss: 1.2311 - val_accuracy: 0.4910
Epoch 12/40
829/829 [==============================] - 11s 14ms/step - loss: 0.7309 - accuracy: 0.6779 - val_loss: 1.2602 - val_accuracy: 0.4729
Epoch 13/40
829/829 [==============================] - 11s 14ms/step - loss: 0.6674 - accuracy: 0.7021 - val_loss: 1.3869 - val_accuracy: 0.5090
Epoch 14/40
829/829 [==============================] - 11s 14ms/step - loss: 0.5873 - accuracy: 0.7648 - val_loss: 2.0446 - val_accuracy: 0.4657
Epoch 15/40
829/829 [==============================] - 11s 14ms/step - loss: 0.5813 - accuracy: 0.7551 - val_loss: 1.8342 - val_accuracy: 0.5018
Epoch 16/40
829/829 [==============================] - 11s 14ms/step - loss: 0.5094 - accuracy: 0.7877 - val_loss: 1.5994 - val_accuracy: 0.5307
Epoch 17/40
829/829 [==============================] - 11s 14ms/step - loss: 0.4369 - accuracy: 0.8275 - val_loss: 2.0702 - val_accuracy: 0.4079
Epoch 18/40
829/829 [==============================] - 11s 14ms/step - loss: 0.3747 - accuracy: 0.8516 - val_loss: 1.7279 - val_accuracy: 0.4657
Epoch 19/40
829/829 [==============================] - 11s 14ms/step - loss: 0.3560 - accuracy: 0.8721 - val_loss: 2.1876 - val_accuracy: 0.5162
Epoch 20/40
829/829 [==============================] - 11s 14ms/step - loss: 0.3717 - accuracy: 0.8456 - val_loss: 1.7713 - val_accuracy: 0.4874
Epoch 21/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2823 - accuracy: 0.8866 - val_loss: 2.7828 - val_accuracy: 0.4079
Epoch 22/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2823 - accuracy: 0.8926 - val_loss: 2.1675 - val_accuracy: 0.4585
Epoch 23/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2486 - accuracy: 0.9059 - val_loss: 1.9805 - val_accuracy: 0.4801
Epoch 24/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2655 - accuracy: 0.9059 - val_loss: 2.0483 - val_accuracy: 0.5054
Epoch 25/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2083 - accuracy: 0.9324 - val_loss: 2.6615 - val_accuracy: 0.4332
Epoch 26/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1949 - accuracy: 0.9312 - val_loss: 2.8827 - val_accuracy: 0.4874
Epoch 27/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1955 - accuracy: 0.9300 - val_loss: 2.4904 - val_accuracy: 0.5162
Epoch 28/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1771 - accuracy: 0.9493 - val_loss: 3.3318 - val_accuracy: 0.4910
Epoch 29/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2101 - accuracy: 0.9361 - val_loss: 3.1583 - val_accuracy: 0.4801
Epoch 30/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2146 - accuracy: 0.9240 - val_loss: 3.4822 - val_accuracy: 0.4801
Epoch 31/40
829/829 [==============================] - 11s 14ms/step - loss: 0.2227 - accuracy: 0.9228 - val_loss: 2.6430 - val_accuracy: 0.4585
Epoch 32/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1474 - accuracy: 0.9493 - val_loss: 2.4249 - val_accuracy: 0.4693
Epoch 33/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1881 - accuracy: 0.9349 - val_loss: 2.9240 - val_accuracy: 0.4729
Epoch 34/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1254 - accuracy: 0.9517 - val_loss: 2.9332 - val_accuracy: 0.4260
Epoch 35/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1566 - accuracy: 0.9409 - val_loss: 3.0209 - val_accuracy: 0.5018
Epoch 36/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1675 - accuracy: 0.9457 - val_loss: 3.3320 - val_accuracy: 0.4982
Epoch 37/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1815 - accuracy: 0.9445 - val_loss: 2.4762 - val_accuracy: 0.4946
Epoch 38/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1494 - accuracy: 0.9433 - val_loss: 3.1968 - val_accuracy: 0.4621
Epoch 39/40
829/829 [==============================] - 11s 14ms/step - loss: 0.0946 - accuracy: 0.9686 - val_loss: 3.6402 - val_accuracy: 0.4368
Epoch 40/40
829/829 [==============================] - 11s 14ms/step - loss: 0.1392 - accuracy: 0.9542 - val_loss: 3.6814 - val_accuracy: 0.4440
history_dict = history.history
loss = history_dict['loss']
val_loss = history_dict['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
acc = history_dict['accuracy']
val_acc = history_dict['val_accuracy']
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation acc')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
model.load_weights('/content/drive/Shared drives/personal/weights/inception_model.16-0.5307.hdf5')
model.evaluate(x_test, y_test)
277/277 [==============================] - 2s 6ms/step
[1.4880577596946745, 0.48736461997032166]
d. 비교를 위한 Fully Connected Neural Network
def build_fcnn():
model = models.Sequential()
model.add(layers.Flatten(input_shape=(256, 256, 3)))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(3, activation='softmax'))
model.compile(optimizer=optimizers.RMSprop(lr=0.0001),
loss='categorical_crossentropy',
metrics=['accuracy'])
return model
model = build_fcnn()
model.name = 'fcnn_model'
model.summary()
Model: "fcnn_model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
flatten_12 (Flatten) (None, 196608) 0
_________________________________________________________________
dense_34 (Dense) (None, 512) 100663808
_________________________________________________________________
dense_35 (Dense) (None, 512) 262656
_________________________________________________________________
dense_36 (Dense) (None, 512) 262656
_________________________________________________________________
dense_37 (Dense) (None, 3) 1539
=================================================================
Total params: 101,190,659
Trainable params: 101,190,659
Non-trainable params: 0
_________________________________________________________________
history = model_train(model, 20)
Train on 829 samples, validate on 277 samples
Epoch 1/20
829/829 [==============================] - 2s 2ms/step - loss: 8.2438 - accuracy: 0.3691 - val_loss: 2.3912 - val_accuracy: 0.4549
Epoch 2/20
829/829 [==============================] - 2s 2ms/step - loss: 3.4840 - accuracy: 0.3848 - val_loss: 2.9575 - val_accuracy: 0.4910
Epoch 3/20
829/829 [==============================] - 2s 2ms/step - loss: 2.1822 - accuracy: 0.3908 - val_loss: 2.1029 - val_accuracy: 0.1697
Epoch 4/20
829/829 [==============================] - 2s 2ms/step - loss: 1.5931 - accuracy: 0.4017 - val_loss: 1.0308 - val_accuracy: 0.4693
Epoch 5/20
829/829 [==============================] - 2s 2ms/step - loss: 1.2660 - accuracy: 0.4427 - val_loss: 1.1608 - val_accuracy: 0.3538
Epoch 6/20
829/829 [==============================] - 2s 2ms/step - loss: 1.1859 - accuracy: 0.4499 - val_loss: 1.1899 - val_accuracy: 0.4404
Epoch 7/20
829/829 [==============================] - 2s 2ms/step - loss: 1.1402 - accuracy: 0.4680 - val_loss: 1.0673 - val_accuracy: 0.4946
Epoch 8/20
829/829 [==============================] - 2s 2ms/step - loss: 1.0462 - accuracy: 0.5078 - val_loss: 1.1018 - val_accuracy: 0.4549
Epoch 9/20
829/829 [==============================] - 2s 2ms/step - loss: 1.0319 - accuracy: 0.4885 - val_loss: 1.1591 - val_accuracy: 0.3682
Epoch 10/20
829/829 [==============================] - 2s 2ms/step - loss: 1.0058 - accuracy: 0.5187 - val_loss: 1.0854 - val_accuracy: 0.4404
Epoch 11/20
829/829 [==============================] - 2s 2ms/step - loss: 1.0152 - accuracy: 0.5115 - val_loss: 1.3047 - val_accuracy: 0.3538
Epoch 12/20
829/829 [==============================] - 2s 2ms/step - loss: 0.9439 - accuracy: 0.5754 - val_loss: 1.1428 - val_accuracy: 0.4007
Epoch 13/20
829/829 [==============================] - 2s 2ms/step - loss: 0.9514 - accuracy: 0.5464 - val_loss: 1.2914 - val_accuracy: 0.4296
Epoch 14/20
829/829 [==============================] - 2s 2ms/step - loss: 0.9071 - accuracy: 0.5862 - val_loss: 1.6851 - val_accuracy: 0.4765
Epoch 15/20
829/829 [==============================] - 2s 2ms/step - loss: 0.8778 - accuracy: 0.6043 - val_loss: 1.2055 - val_accuracy: 0.3610
Epoch 16/20
829/829 [==============================] - 2s 2ms/step - loss: 0.8328 - accuracy: 0.6188 - val_loss: 1.6090 - val_accuracy: 0.3755
Epoch 17/20
829/829 [==============================] - 2s 2ms/step - loss: 0.8415 - accuracy: 0.6454 - val_loss: 1.5808 - val_accuracy: 0.4332
Epoch 18/20
829/829 [==============================] - 2s 2ms/step - loss: 0.7492 - accuracy: 0.6779 - val_loss: 1.5207 - val_accuracy: 0.3935
Epoch 19/20
829/829 [==============================] - 2s 2ms/step - loss: 0.7720 - accuracy: 0.6876 - val_loss: 1.4689 - val_accuracy: 0.4224
Epoch 20/20
829/829 [==============================] - 2s 2ms/step - loss: 0.7792 - accuracy: 0.6948 - val_loss: 1.2987 - val_accuracy: 0.4260
history_dict = history.history
loss = history_dict['loss']
val_loss = history_dict['val_loss']
epochs = range(1, len(loss) + 1)
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
acc = history_dict['accuracy']
val_acc = history_dict['val_accuracy']
plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation acc')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
model.load_weights('/content/drive/Shared drives/personal/weights/fcnn_model.07-0.4946.hdf5')
model.evaluate(x_test, y_test)
277/277 [==============================] - 0s 520us/step
[1.0848816297544899, 0.45126354694366455]