def load_data():
"""Get data with labels, split into training, validation and test set."""
ourtrainingdata = get_training_data()
ourtestingdata = get_testing_data()
#X_train, y_train = data[0]
X_train = ourtrainingdata[0]
y_train = ourtrainingdata[1]
#X_test, y_test = data[2]
X_test = ourtestingdata[0]
y_test = ourtestingdata[1]
y_train = numpy.asarray(y_train, dtype=numpy.int32)
y_test = numpy.asarray(y_test, dtype=numpy.int32)
return dict(
X_train=X_train,
y_train=y_train,
X_test=X_test,
y_test=y_test,
num_examples_train=len(X_train),
num_examples_test=len(X_test),
input_dim=X_train[0],
output_dim=10,
)
from prepare_data import get_training_data
from prepare_plots import plot_results
from build_cnn_encoderdecoder_model import build_cnn_encoderdecoder_model
import numpy as np
if __name__ == "__main__":
profile_pngs_objs, midcurve_pngs_objs = get_training_data(size=(128, 128))
profile_pngs_objs = np.asarray(profile_pngs_objs)
midcurve_pngs_objs = np.asarray(midcurve_pngs_objs)
profile_pngs_objs = np.expand_dims(profile_pngs_objs, axis=-1)
midcurve_pngs_objs = np.expand_dims(midcurve_pngs_objs, axis=-1)
profile_pngs_objs = (profile_pngs_objs - 127.5) / 127.5 #Normalize [-1, 1]
midcurve_pngs_objs = (midcurve_pngs_objs -
127.5) / 127.5 #Normalize [-1, 1]
x_coord = np.zeros(shape=(128, 128, 1))
y_coord = np.zeros(shape=(128, 128, 1))
for i in range(0, 128):
x_coord[:, i, 0] = i
y_coord[i, :, 0] = i
coords = np.append(x_coord, y_coord, axis=-1)
coords = (coords - 63.5) / 63.5 #Normalize [-1, 1]
#coords = np.expand_dims(coords, axis=0)
profile_pngs = []
for i in range(len(profile_pngs_objs)):
self.x = profile_pngs_objs
self.y = midcurve_pngs_objs
self.autoencoder.fit(self.x, self.y,
epochs=self.epochs,
batch_size=5,
shuffle=True)
# Save models
self.autoencoder.save(self.autoencoder_model_pkl)
self.encoder.save(self.encoder_model_pkl)
self.decoder.save(self.decoder_model_pkl)
else:
# Save models
self.autoencoder = load_model(self.autoencoder_model_pkl)
self.encoder= load_model(self.encoder_model_pkl)
self.decoder = load_model(self.decoder_model_pkl)
def predict(self, test_profile_images):
png_profile_images = self.process_images(test_profile_images)
encoded_imgs = self.encoder.predict(png_profile_images)
decoded_imgs = self.decoder.predict(encoded_imgs)
return test_profile_images,decoded_imgs
if __name__ == "__main__":
profile_gray_objs, midcurve_gray_objs = get_training_data()
endec = simple_encoderdecoder()
endec.train(profile_gray_objs, midcurve_gray_objs)
test_gray_images = random.sample(profile_gray_objs,5)
original_profile_imgs,predicted_midcurve_imgs = endec.predict(test_gray_images)
plot_results(original_profile_imgs,predicted_midcurve_imgs)
data_dir)
num_classes = len(set(train_labels))
print(num_classes)
# Build the graph using appropriate class from slim
graph = tf.Graph()
ckpt = tf.train.latest_checkpoint(os.getcwd())
with graph.as_default():
if ckpt == None:
# Indicates whether we are in training or in test mode
is_training = tf.placeholder(tf.bool, name="is_training")
# Get data iterators
images_train, labels_train, train_init_op, iterator = pd.get_training_data(
train_filenames, train_labels)
images_val, labels_val, val_init_op, _ = pd.get_validation_data(
val_filenames, val_labels, iterator_val=iterator)
# Get the pretrained model, specifying the num_classes argument to create a new
# fully connected replacing the last one.
# Each model has a different architecture, so the final layer will change in another model.
# Here, logits gives us directly the predicted scores we wanted from the images.
logits = dm.predict(model_type, images_train, num_classes,
weight_decay, drop_out, True)
# Specify where the model checkpoint is (pretrained weights).
assert (os.path.isfile(model_path))
# Restore layers that we want
from prepare_data import get_training_data
from prepare_plots import plot_results
from build_simple_encoderdecoder_w_denoiser_model import simple_encoderdecoder_w_denoiser
if __name__ == "__main__":
profile_pngs_objs, midcurve_pngs_objs = get_training_data()
endec = simple_encoderdecoder_w_denoiser()
original_imgs,decoded_imgs = endec.simple_encoderdecoder_w_denoiser(profile_pngs_objs, midcurve_pngs_objs)
plot_results(original_imgs,decoded_imgs)