def train_top_model():
train_data = np.load(
os.path.join(result_dir, 'bottleneck_features_train.npy'))
train_labels = np.array([0] * int(nb_train_samples / 2) +
[1] * int(nb_train_samples / 2))
print(train_data.shape)
validation_data = np.load(
os.path.join(result_dir, 'bottleneck_features_validation.npy'))
validation_labels = np.array([0] * int(nb_validation_samples / 2) +
[1] * int(nb_validation_samples / 2))
print(validation_data.shape)
model = Sequential()
model.add(Flatten(input_shape=train_data.shape[1:]))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy',
optimizer=optimizers.SGD(lr=1e-4, momentum=0.9),
metrics=['accuracy'])
history = model.fit(train_data,
train_labels,
nb_epoch=nb_epoch,
batch_size=32,
validation_data=(validation_data, validation_labels))
model.save_weights(os.path.join(result_dir, 'bottleneck_fc_model.h5'))
save_history(history, os.path.join(result_dir, 'history_extractor.txt'))
def train_top_model():
"""VGGのボトルネック特徴量を入力とし、Dog vs Catの正解を出力とするFCネットワークを訓練"""
# 訓練データをロード
# ジェネレータではshuffle=Falseなので最初の1000枚がcats、次の1000枚がdogs
train_data = np.load(
os.path.join(result_dir, 'bottleneck_features_train.npy'))
train_labels = np.array([0] * int(nb_train_samples / 2) +
[1] * int(nb_train_samples / 2))
# (2000, 4, 4, 512)
print(train_data.shape)
# バリデーションデータをロード
validation_data = np.load(
os.path.join(result_dir, 'bottleneck_features_validation.npy'))
validation_labels = np.array([0] * int(nb_validation_samples / 2) +
[1] * int(nb_validation_samples / 2))
# (800, 4, 4, 512)
print(validation_data.shape)
# FCネットワークを構築
model = Sequential()
model.add(Flatten(input_shape=train_data.shape[1:]))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy',
optimizer=optimizers.SGD(lr=1e-4, momentum=0.9),
metrics=['accuracy'])
history = model.fit(train_data,
train_labels,
nb_epoch=nb_epoch,
batch_size=32,
validation_data=(validation_data, validation_labels))
model.save_weights(os.path.join(result_dir, 'bottleneck_fc_model.h5'))
save_history(history, os.path.join(result_dir, 'history_extractor.txt'))
def train_top_model():
"""VGGのボトルネック特徴量を入力とし、正解を出力とするFCネットワークを訓練"""
# 訓練データをロード
# ジェネレータではshuffle=Falseなのでクラスは順番に出てくる
# one-hot vector表現へ変換が必要
train_data = np.load(
os.path.join(result_dir, 'bottleneck_features_train.npy'))
train_labels = [i // nb_samples_per_class for i in range(nb_train_samples)]
train_labels = np_utils.to_categorical(train_labels, nb_classes)
# バリデーションデータをロード
validation_data = np.load(
os.path.join(result_dir, 'bottleneck_features_validation.npy'))
validation_labels = [
i // nb_samples_per_class for i in range(nb_val_samples)
]
validation_labels = np_utils.to_categorical(validation_labels, nb_classes)
# FCネットワークを構築
model = Sequential()
model.add(Flatten(input_shape=train_data.shape[1:]))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes, activation='softmax'))
model.summary()
model.compile(loss='categorical_crossentropy',
optimizer=optimizers.SGD(lr=1e-4, momentum=0.9),
metrics=['accuracy'])
history = model.fit(train_data,
train_labels,
nb_epoch=nb_epoch,
batch_size=batch_size,
validation_data=(validation_data, validation_labels))
model.save_weights(os.path.join(result_dir, 'bottleneck_fc_model.h5'))
save_history(history, os.path.join(result_dir, 'history_extractor.txt'))
preprocessing_function=preprocess_input)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_rows, img_cols),
color_mode='rgb',
classes=classes,
class_mode='categorical',
batch_size=batch_size,
shuffle=True)
validation_generator = test_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_rows, img_cols),
color_mode='rgb',
classes=classes,
class_mode='categorical',
batch_size=batch_size,
shuffle=True)
# Fine-tuning
history = model.fit_generator(
train_generator,
samples_per_epoch=nb_train_samples,
nb_epoch=nb_epoch,
validation_data=validation_generator,
nb_val_samples=nb_val_samples)
model.save_weights(os.path.join(result_dir, 'finetuning.h5'))
save_history(history, os.path.join(result_dir, 'history_finetuning.txt'))
test_datagen = ImageDataGenerator(rescale=1.0 / 255)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_rows, img_cols),
color_mode='rgb',
classes=classes,
class_mode='categorical',
batch_size=batch_size,
shuffle=True)
validation_generator = test_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_rows, img_cols),
color_mode='rgb',
classes=classes,
class_mode='categorical',
batch_size=batch_size,
shuffle=True)
# モデル訓練
history = model.fit_generator(train_generator,
samples_per_epoch=nb_train_samples,
nb_epoch=nb_epoch,
validation_data=validation_generator,
nb_val_samples=nb_val_samples)
model.save_weights(os.path.join(result_dir, 'vgg_scratch.h5'))
save_history(history, os.path.join(result_dir, 'history_vgg_scratch.txt'))