def evaluate(self):
train_generator, val_generator = get_data(batch_size=batch_size)
model_path = os.path.join(log_dir, 'mobilenetv1.h5')
model = self.mobilenet_v1()
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
model.evaluate(val_generator)
def train(self):
epochs = 100
np.random.seed(200)
model = self.xception()
batch_size = 64
train_generator, val_generator = get_data(batch_size=batch_size)
if not os.path.exists(log_dir):
os.system('mkdir -p {}'.format(log_dir))
model_path = os.path.join(log_dir, 'xception_reluswish.h5')
callbacks = [
keras.callbacks.TensorBoard(log_dir),
keras.callbacks.ModelCheckpoint(model_path, save_best_only=True),
]
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples // batch_size,
epochs=epochs,
validation_data=val_generator,
validation_steps=val_generator.samples // batch_size,
callbacks=callbacks)
def train(self):
epochs = 500
model = self.alexnet()
batch_size = 220
train_generator, val_generator = get_data(batch_size=batch_size)
if not os.path.exists(log_dir):
os.system('mkdir -p {}'.format(log_dir))
else:
os.system('rm -rf {}'.format(log_dir))
model_path = os.path.join(log_dir, 'alexnet_relu6.h5')
callbacks = [
keras.callbacks.TensorBoard(log_dir),
keras.callbacks.ModelCheckpoint(model_path, save_best_only=True),
]
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples // batch_size,
epochs=epochs,
validation_data=val_generator,
validation_steps=val_generator.samples // batch_size,
callbacks=callbacks)
def evaluate(self):
batch_size = 256
train_generator, val_generator = get_data(batch_size=batch_size)
model_path = os.path.join(log_dir, 'alexnet_reluswish.h5')
model = self.alexnet()
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
model.evaluate(val_generator)
def train(self):
epochs = 100
model = self.vgg16()
batch_size = 8
train_generator, val_generator = get_data(batch_size=batch_size)
model_path = os.path.join(log_dir, 'vgg16.h5')
callbacks = [
keras.callbacks.TensorBoard(log_dir),
keras.callbacks.ModelCheckpoint(model_path, save_best_only=True),
]
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
history = model.fit_generator(train_generator, steps_per_epoch=train_generator.samples//batch_size,
epochs=epochs,
validation_data=val_generator,
validation_steps=val_generator.samples//batch_size,
callbacks=callbacks)
def train(self):
epochs = 10
np.random.seed(200)
model = self.my_model()
batch_size = 32
train_generator, val_generator = get_data(batch_size=batch_size)
model_path = os.path.join(log_dir, 'mymodel.h5')
callbacks = [
keras.callbacks.TensorBoard(log_dir),
keras.callbacks.ModelCheckpoint(model_path),
]
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples // batch_size,
epochs=epochs,
validation_data=val_generator,
validation_steps=val_generator.samples // batch_size,
callbacks=callbacks)
def train(self):
epochs = 1
np.random.seed(200)
model = self.thrid_party_resnext_50()
batch_size = 16
train_generator, val_generator = get_data(batch_size=batch_size)
model_path = os.path.join(log_dir, 'resnext50.h5')
callbacks = [
keras.callbacks.TensorBoard(log_dir),
keras.callbacks.ModelCheckpoint(model_path, save_best_only=True),
]
if os.path.exists(model_path):
model = keras.models.load_model(model_path)
history = model.fit_generator(
train_generator,
steps_per_epoch=train_generator.samples // batch_size,
epochs=epochs,
validation_data=val_generator,
validation_steps=val_generator.samples // batch_size,
)
# callbacks=callbacks)
model.summary()
