def on_epoch_end(self, epoch, logs=None):
if logs is not None:
if logs['val_loss'] < self.max_val_loss:
self.max_val_loss = logs['val_loss']
weight_file_name = 'model_weights_best'
model_tools.save_network(self.model, weight_file_name)
np.save('lossRecord', self.max_val_loss)
print("best weights and min loss saved")
def train_model(learning_rate, batch_size, num_epochs, steps_per_epoch,
validation_steps, workers, optimizer, weight_file_name):
image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3
# Call fcn_model()
output_layer = fcn_model(inputs, num_classes)
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)
if optimizer == 'Nadam':
model.compile(optimizer=keras.optimizers.Adam(learning_rate),
loss='categorical_crossentropy')
elif optimizer == 'Adam':
model.compile(optimizer=keras.optimizers.Nadam(lr=learning_rate,
beta_1=0.9,
beta_2=0.999,
epsilon=1e-08,
schedule_decay=0.004),
loss='categorical_crossentropy')
else:
raise RuntimeError('Invalid optimizer type "{}"'.format(optimizer))
# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
data_folder=os.path.join(
'..', 'data', 'train'),
image_shape=image_shape,
shift_aug=True)
val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
data_folder=os.path.join(
'..', 'data',
'validation'),
image_shape=image_shape)
logger_cb = plotting_tools.LoggerPlotter(save_graphs=True)
callbacks = [logger_cb]
model.fit_generator(
train_iter,
steps_per_epoch=steps_per_epoch, # the number of batches per epoch,
epochs=num_epochs, # the number of epochs to train for,
validation_data=val_iter, # validation iterator
validation_steps=
validation_steps, # the number of batches to validate on
callbacks=callbacks,
workers=workers)
# Save your trained model weights
model_tools.save_network(model, weight_file_name)
#logger_cb = plotting_tools.LoggerPlotter()
#callbacks = [logger_cb]
callbacks = []
model.fit_generator(
train_iter,
steps_per_epoch=steps_per_epoch, # the number of batches per epoch,
epochs=num_epochs, # the number of epochs to train for,
validation_data=val_iter, # validation iterator
validation_steps=
validation_steps, # the number of batches to validate on
callbacks=callbacks,
workers=workers)
weight_file_name = 'model_weights_' + run
model_tools.save_network(model, weight_file_name)
run_num = 'run_' + run
val_with_targ, pred_with_targ = model_tools.write_predictions_grade_set(
model, run_num, 'patrol_with_targ', 'evaluation')
val_no_targ, pred_no_targ = model_tools.write_predictions_grade_set(
model, run_num, 'patrol_non_targ', 'evaluation')
val_following, pred_following = model_tools.write_predictions_grade_set(
model, run_num, 'following_images', 'evaluation')
print(
"--------------------------------------------------------------------------------"
)
val_with_targ, pred_with_targ)
# Sum all the true positives, etc from the three datasets to get a weight for the score
true_pos = true_pos1 + true_pos2 + true_pos3
false_pos = false_pos1 + false_pos2 + false_pos3
false_neg = false_neg1 + false_neg2 + false_neg3
weight = true_pos / (true_pos + false_neg + false_pos)
#print(weight)
# The IoU for the dataset that never includes the hero is excluded from grading
final_IoU = (iou1 + iou3) / 2
print("IoU no hero - ", final_IoU)
# And the final grade score is
final_score = final_IoU * weight
print("Final Grade - ", final_score)
if final_score > fg:
# Save model with best score
weight_file_name = str(final_score) + '_' + datetime.now(
).strftime('model_%H_%M_%d_%m_%Y') + '.h5'
model_tools.save_network(your_model=model,
your_weight_filename=weight_file_name)
fg = final_score
arr = [fg, final_IoU, weight, weight_file_name]
print('Best --- ', fg)
print('Data --- ', arr)
def train_net(x):
"""Trains the network with the suggested hyper-parameters passed in x
argument. Returns -final_score as a result to let ski-opt library find the
desired minimum, corresponding to the maximum final_score"""
learning_rate = x[0] # 0.001
batch_size = x[1] # 32
num_epochs = x[2] # 12
layers_num = x[3] #2
conv_layers_num = x[4] #1
external_features = x[5] #128
internal_features = x[6] #16
conv_features = x[7] #16
print()
print("learning_rate", learning_rate)
print("batch_size", batch_size)
print("num_epochs", num_epochs)
print("layers_num", layers_num)
print("conv_layers_num", conv_layers_num)
print("external_features", external_features)
print("internal_features", internal_features)
print("conv_features", conv_features)
image_hw = 160
image_shape = (image_hw, image_hw, 3)
inputs = layers.Input(image_shape)
num_classes = 3
# Call fcn_model()
output_layer = fcn_model(inputs, num_classes, layers_num,
external_features, internal_features,
conv_layers_num, conv_features)
# Define the Keras model and compile it for training
model = models.Model(inputs=inputs, outputs=output_layer)
model.compile(optimizer=keras.optimizers.Adam(learning_rate),
loss='categorical_crossentropy')
# Data iterators for loading the training and validation data
train_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
data_folder=os.path.join(
'..', 'data',
'train_combined'),
image_shape=image_shape,
shift_aug=True)
val_iter = data_iterator.BatchIteratorSimple(batch_size=batch_size,
data_folder=os.path.join(
'..', 'data',
'validation'),
image_shape=image_shape)
model.fit_generator(
train_iter,
steps_per_epoch=steps_per_epoch, # the number of batches per epoch,
epochs=num_epochs, # the number of epochs to train for,
validation_data=val_iter, # validation iterator
validation_steps=
validation_steps, # the number of batches to validate on
workers=workers)
run_num = 'run_1'
val_with_targ, pred_with_targ = model_tools.write_predictions_grade_set(
model, run_num, 'patrol_with_targ', 'sample_evaluation_data')
val_no_targ, pred_no_targ = model_tools.write_predictions_grade_set(
model, run_num, 'patrol_non_targ', 'sample_evaluation_data')
val_following, pred_following = model_tools.write_predictions_grade_set(
model, run_num, 'following_images', 'sample_evaluation_data')
# Scores for while the quad is following behind the target.
true_pos1, false_pos1, false_neg1, iou1 = scoring_utils.score_run_iou(
val_following, pred_following)
# Scores for images while the quad is on patrol and the target is not
# visible
true_pos2, false_pos2, false_neg2, iou2 = scoring_utils.score_run_iou(
val_no_targ, pred_no_targ)
# This score measures how well the neural network can detect the target
# from far away
true_pos3, false_pos3, false_neg3, iou3 = scoring_utils.score_run_iou(
val_with_targ, pred_with_targ)
# Sum all the true positives, etc from the three datasets to get a weight
# for the score
true_pos = true_pos1 + true_pos2 + true_pos3
false_pos = false_pos1 + false_pos2 + false_pos3
false_neg = false_neg1 + false_neg2 + false_neg3
weight = true_pos / (true_pos + false_neg + false_pos)
# The IoU for the dataset that never includes the hero is excluded from
# grading
final_IoU = (iou1 + iou3) / 2
# And the final grade score is
final_score = final_IoU * weight
weight_file_name = 'model_weights_' + str(final_score)
model_tools.save_network(model, weight_file_name)
print("Saved", weight_file_name)
print("final_score", final_score)
print()
return -final_score