def read_image(self, filename):
target_size = (self.configuration.get_image_height(),
self.configuration.get_image_width())
image = self.process_fun(
data.read_image(filename,
self.configuration.get_number_of_channels()),
target_size)
return image
validation_steps=configuration.get_validation_steps(),
callbacks=[model_checkpoint_callback],
)
elif pargs.mode == "test":
model.evaluate(val_dataset, steps=configuration.get_validation_steps())
elif pargs.mode == "predict":
filename = input("file :")
while filename != "end":
target_size = (
configuration.get_image_height(),
configuration.get_image_width(),
)
image = process_fun(
data.read_image(filename,
configuration.get_number_of_channels()),
target_size,
)
image = image - mean_image
image = tf.expand_dims(image, 0)
pred = model.predict(image)
pred = pred[0]
# softmax to estimate probs
pred = np.exp(pred - max(pred))
pred = pred / np.sum(pred)
cla = np.argmax(pred)
print(pred)
print(cla)
filename = input("file :")
# save the model
if pargs.save:
if pargs.mode == 'train' :
history = model.fit(tr_dataset,
epochs = configuration.get_number_of_epochs(),
validation_data=val_dataset,
validation_steps = configuration.get_validation_steps(),
callbacks=[model_checkpoint_callback])
elif pargs.mode == 'test' :
model.evaluate(val_dataset,
steps = configuration.get_validation_steps())
elif pargs.mode == 'predict':
filename = input('file :')
while(filename != 'end') :
target_size = (configuration.get_image_height(), configuration.get_image_width())
image = process_fun(data.read_image(filename, configuration.get_number_of_channels()), target_size )
image = image - mean_image
image = tf.expand_dims(image, 0)
pred = model.predict(image)
pred = pred[0]
#softmax to estimate probs
pred = np.exp(pred - max(pred))
pred = pred / np.sum(pred)
cla = np.argmax(pred)
print(pred)
print(cla)
filename = input('file :')
#save the model
if pargs.save :
saved_to = os.path.join(configuration.get_data_dir(),"cnn-model")
model.save(saved_to)