def predict_command(model, training_cnf, features_file, images_dir, weights_from, tag, sync, ):
util.check_required_program_args([model, training_cnf, features_file, images_dir, weights_from])
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
weights_from = str(weights_from)
image_features = np.load(features_file)
images = data.get_image_files(images_dir)
predictions = predict_withf(model, cnf, weights_from, image_features)
predict_dir = os.path.dirname(features_file)
prediction_results_dir = os.path.abspath(os.path.join(predict_dir, 'predictions', tag))
if not os.path.exists(prediction_results_dir):
os.makedirs(prediction_results_dir)
names = data.get_names(images)
image_prediction_probs = np.column_stack([names, predictions])
headers = ['score%d' % (i + 1) for i in range(predictions.shape[1])]
title = np.array(['image'] + headers)
image_prediction_probs = np.vstack([title, image_prediction_probs])
prediction_probs_file = os.path.join(prediction_results_dir, 'predictions.csv')
np.savetxt(prediction_probs_file, image_prediction_probs, delimiter=",", fmt="%s")
print('Predictions saved to: %s' % prediction_probs_file)
if cnf['classification']:
class_predictions = np.argmax(predictions, axis=1)
image_class_predictions = np.column_stack([names, class_predictions])
title = np.array(['image', 'label'])
image_class_predictions = np.vstack([title, image_class_predictions])
prediction_class_file = os.path.join(prediction_results_dir, 'predictions_class.csv')
np.savetxt(prediction_class_file, image_class_predictions, delimiter=",", fmt="%s")
print('Class predictions saved to: %s' % prediction_class_file)
def main(model, training_cnf, data_dir, parallel, start_epoch, weights_from,
resume_lr, gpu_memory_fraction, is_summary, num_classes):
model_def = util.load_module(model)
model = model_def
cnf = util.load_module(training_cnf).cnf
util.init_logging('train_ss.log',
file_log_level=logging.INFO,
console_log_level=logging.INFO)
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
standardizer = cnf.get('standardizer', NoOpStandardizer())
training_iter, validation_iter = create_training_iters(cnf,
data_set,
standardizer,
model_def.crop_size,
start_epoch,
parallel=parallel)
trainer = GenerativeLearner(model,
cnf,
training_iterator=training_iter,
validation_iterator=validation_iter,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
is_summary=is_summary,
verbosity=2)
trainer.fit(data_set,
num_classes,
weights_from,
start_epoch,
summary_every=399)
def main(model, training_cnf, data_dir, parallel, start_epoch, task_id, job_name, ps_hosts, worker_hosts, weights_from, resume_lr, gpu_memory_fraction, is_summary, loss_type):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
ps_hosts = ps_hosts.split(',')
worker_hosts = worker_hosts.split(',')
cluster_spec = tf.train.ClusterSpec({'ps': ps_hosts,
'worker': worker_hosts})
server = tf.train.Server(
{'ps': ps_hosts,
'worker': worker_hosts},
job_name=job_name,
task_index=task_id)
util.init_logging('train.log', file_log_level=logging.INFO,
console_log_level=logging.INFO)
if weights_from:
weights_from = str(weights_from)
if job_name == 'ps':
server.join()
else:
learner = DistSupervisedLearner(model, cnf, resume_lr=resume_lr, classification=cnf[
'classification'], gpu_memory_fraction=gpu_memory_fraction, is_summary=is_summary, loss_type=loss_type, verbosity=1)
data_dir_train = os.path.join(data_dir, 'train')
data_dir_val = os.path.join(data_dir, 'val')
learner.fit(task_id, server, cluster_spec, data_dir_train, data_dir_val, weights_from=weights_from, start_epoch=start_epoch, training_set_size=50000, val_set_size=10000,
summary_every=399, keep_moving_averages=True)
def predict(model, training_cnf, predict_dir, weights_from, predict_type):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
weights_from = str(weights_from)
images = data.get_image_files(predict_dir)
preprocessor = None
prediction_iterator = create_prediction_iter(cnf, model_def.crop_size,
preprocessor, False)
if predict_type == 'quasi':
predictor = QuasiCropPredictor(model, cnf, weights_from,
prediction_iterator, 20)
elif predict_type == '1_crop':
predictor = OneCropPredictor(model, cnf, weights_from,
prediction_iterator)
elif predict_type == '10_crop':
predictor = TenCropPredictor(model, cnf, weights_from,
prediction_iterator,
model_def.crop_size[0],
model_def.image_size[0])
else:
raise ValueError('Unknown predict_type: %s' % predict_type)
predictions = predictor.predict(images)
predictions = predictions.reshape(-1, 1000)
names = data.get_names(images)
for i, name in enumerate(names):
print("---Predictions for %s:" % name)
preds = (np.argsort(predictions[i])[::-1])[0:5]
for p in preds:
print(class_names[p], predictions[i][p])
def main(model, training_cnf, data_dir, start_epoch, resume_lr, weights_from,
clean, visuals):
util.check_required_program_args([model, training_cnf, data_dir])
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
util.init_logging('train.log',
file_log_level=logging.INFO,
console_log_level=logging.INFO,
clean=clean)
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
training_iter, validation_iter = create_training_iters(
cnf, data_set, model_def.crop_size, start_epoch,
cnf.get('iterator_type', 'queued') == 'parallel')
trainer = SupervisedTrainer(model,
cnf,
training_iter,
validation_iter,
classification=cnf['classification'])
trainer.fit(data_set,
weights_from,
start_epoch,
resume_lr,
verbose=1,
summary_every=cnf.get('summary_every', 10),
clean=clean,
visuals=visuals)
def main(model, training_cnf, data_dir, parallel, max_to_keep, start_epoch,
weights_from, weights_dir, num_classes, resume_lr,
gpu_memory_fraction, is_summary, loss_type, weighted, data_balancing,
log_file_name):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
if weights_from:
weights_from = str(weights_from)
learner = SupervisedLearner(model,
cnf,
data_balancing=data_balancing,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
num_classes=num_classes,
is_summary=is_summary,
loss_type=loss_type,
verbosity=1,
weighted=weighted,
log_file_name=log_file_name)
data_dir_train = os.path.join(data_dir, 'train')
data_dir_val = os.path.join(data_dir, 'val')
learner.fit(data_dir_train,
data_dir_val,
weights_from=weights_from,
weights_dir=weights_dir,
max_to_keep=max_to_keep,
start_epoch=start_epoch,
training_set_size=50000,
val_set_size=10000,
summary_every=399,
keep_moving_averages=True)
def predict(model, training_cnf, predict_dir, weights_from, dataset_name,
convert, image_size, sync, test_type, gpu_memory_fraction):
model = util.load_module(model)
cnf = util.load_module(training_cnf).cnf
weights_from = str(weights_from)
with tf.Graph().as_default():
end_points_G = model.generator([32, 100], True, None)
inputs = tf.placeholder(tf.float32,
shape=(None, model.image_size[0],
model.image_size[0], 3),
name="input")
end_points_D = model.discriminator(inputs,
True,
None,
num_classes=6,
batch_size=32)
saver = tf.train.Saver()
print('Loading weights from: %s' % weights_from)
if gpu_memory_fraction is not None:
gpu_options = tf.GPUOptions(
per_process_gpu_memory_fraction=gpu_memory_fraction)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
saver.restore(sess, weights_from)
end_points_G_val = model.generator([cnf['batch_size_test'], 100],
False,
True,
batch_size=cnf['batch_size_test'])
util.save_images('generated_images.png',
sess.run(end_points_G_val['softmax']),
width=128,
height=128)
sess.close()
def load_model(model, model_cnf, weights_from, layer_name, data_path):
model_def = util.load_module(model)
cnf = util.load_module(model_cnf).cnf
standardizer = cnf['standardizer']
model = model_def.model
sess = tf.Session()
try:
end_points = model(is_training=False, reuse=None)
saver = tf.train.Saver()
print('Loading weights from: %s' % weights_from)
saver.restore(sess, weights_from)
except Exception:
print('not loaded')
inputs = end_points['inputs']
layer = end_points[layer_name]
data = tf.read_file(data_path)
data = tf.to_float(tf.image.decode_jpeg(data))
data = standardizer(data, False)
data = tf.transpose(data, perm=[1, 0, 2])
data_ = data.eval(session=sess)
data_ = scipy.misc.imresize(data_, size=(448, 448), interp='cubic')
data_ = np.expand_dims(data_, 0)
acti = get_activation(sess, layer, inputs, data_)
acti = np.mean(acti, 3).squeeze()
# acti = np.asarray(acti.transpose(1, 0), dtype=np.float32)
plt.imshow(acti)
plt.show()
# print(acti)
"""
def main(model, training_cnf, data_dir, parallel, start_epoch, weights_from,
weights_dir, resume_lr, num_classes, gpu_memory_fraction, is_summary,
loss_type, log_file_name):
with tf.Graph().as_default():
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
if weights_from:
weights_from = str(weights_from)
trainer = SupervisedLearner(model,
cnf,
log_file_name=log_file_name,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
num_classes=num_classes,
is_summary=is_summary,
loss_type=loss_type,
verbosity=1)
trainer.fit(data_dir,
weights_from=weights_from,
weights_dir=weights_dir,
start_epoch=start_epoch,
summary_every=399,
keep_moving_averages=True)
def main(model, training_cnf, data_dir, parallel, start_epoch, weights_from,
resume_lr, gpu_memory_fraction, is_summary, num_classes,
log_file_name):
model_def = util.load_module(model)
model = model_def
cnf = util.load_module(training_cnf).cnf
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
standardizer = cnf.get('standardizer', NoOpStandardizer())
training_iter, validation_iter = create_training_iters(cnf,
data_set,
standardizer,
model_def.crop_size,
start_epoch,
parallel=parallel)
trainer = SemiSupervisedTrainer(model,
cnf,
training_iterator=training_iter,
validation_iterator=validation_iter,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
is_summary=is_summary,
verbosity=1,
log_file_name=log_file_name)
trainer.fit(data_set,
num_classes,
weights_from,
start_epoch,
summary_every=399)
def main(model, training_cnf, data_dir, parallel, start_epoch, weights_from,
resume_lr, gpu_memory_fraction, is_summary):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
util.init_logging('train.log',
file_log_level=logging.INFO,
console_log_level=logging.INFO)
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
standardizer = cnf.get('standardizer', NoOpStandardizer())
training_iter, validation_iter = create_training_iters(cnf,
data_set,
standardizer,
model_def.crop_size,
start_epoch,
parallel=parallel)
trainer = SupervisedTrainer(model,
cnf,
training_iter,
validation_iter,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
is_summary=is_summary,
loss_type='kappa_log')
trainer.fit(data_set,
weights_from,
start_epoch,
verbose=1,
summary_every=399)
def main(model, training_cnf, data_dir, start_epoch, resume_lr, weights_from, clean):
util.check_required_program_args([model, training_cnf, data_dir])
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
util.init_logging('train.log', file_log_level=logging.INFO, console_log_level=logging.INFO, clean=clean)
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
training_iter = BatchIterator(cnf['batch_size_train'], True)
validation_iter = BatchIterator(cnf['batch_size_test'], True)
trainer = SupervisedTrainer(model, cnf, training_iter, validation_iter, classification=cnf['classification'])
trainer.fit(data_set, weights_from, start_epoch, resume_lr, verbose=1,
summary_every=cnf.get('summary_every', 10), clean=clean)
def predict(frozen_model, training_cnf, predict_dir, image_size, output_path,
num_classes, gpu_memory_fraction):
cnf = util.load_module(training_cnf).cnf
standardizer = cnf['standardizer']
graph = util.load_frozen_graph(frozen_model)
preprocessor = convert_preprocessor(image_size)
predictor = SegmentPredictor(graph, standardizer, preprocessor)
# images = data.get_image_files(predict_dir)
image_names = [
filename.strip() for filename in os.listdir(predict_dir)
if filename.endswith('.jpg')
]
hist = np.zeros((num_classes, num_classes))
for image_filename in image_names:
final_prediction_map = predictor.predict(
os.path.join(predict_dir, image_filename))
final_prediction_map = final_prediction_map.transpose(0, 2,
1).squeeze()
gt_name = os.path.join(predict_dir,
image_filename[:-4] + '_final_mask' + '.png')
gt = convert(gt_name, image_size)
gt = np.asarray(gt)
gt = convert_labels(gt, image_size, image_size)
hist += compute_hist(gt, final_prediction_map, num_classes=num_classes)
iou = np.diag(hist) / (hist.sum(1) + hist.sum(0) - np.diag(hist))
meaniou = np.nanmean(iou)
print('Mean IOU %5.5f' % meaniou)
def __init__(self,
model_cnf,
data_dir,
image_size,
crop_size,
channel_dim=3,
start_epoch=1,
parallel=True):
self.cnf = util.load_module(model_cnf).cnf
self.image_size = image_size
self.crop_size = crop_size
self.channel_dim = channel_dim
self.data_set = DataSet(data_dir, image_size)
self.standardizer = self.cnf.get('standardizer', NoOpStandardizer())
self.training_iter, self.validation_iter = create_training_iters(
self.cnf,
self.data_set,
self.standardizer, [crop_size, crop_size],
start_epoch,
parallel=parallel)
self.training_X = self.data_set.training_X
self.training_y = self.data_set.training_y
self.validation_X = self.data_set.validation_X
self.validation_y = self.data_set.validation_y
def model_info(model):
util.check_required_program_args([model])
model_def = util.load_module(model)
model = model_def.model
end_points = model(False, None)
util.show_layer_shapes(end_points)
util.show_vars()
def try_config(args, cnf):
"""For trying out configurations.
Args:
args: command line arguments regarding training
cnf: training configuration sampled from hyperband search space
Returns:
a dictionary containing final loss value and early stop flag
"""
model_def = util.load_module(args['model'])
model = model_def.model
if args['weights_from']:
weights_from = str(args['weights_from'])
else:
weights_from = args['weights_from']
data_set = DataSet(args['data_dir'],
model_def.image_size[0],
mode=cnf.get('mode'),
multilabel=cnf.get('multilabel', False))
standardizer = cnf.get('standardizer', NoOpStandardizer())
cutout = cnf.get('cutout', None)
training_iter, validation_iter = create_training_iters(
cnf,
data_set,
standardizer,
model_def.crop_size,
args['start_epoch'],
parallel=args['parallel'],
cutout=cutout,
data_balancing=cnf.get('data_balancing', False))
learner = SupervisedLearner(
model,
cnf,
training_iterator=training_iter,
validation_iterator=validation_iter,
resume_lr=args['resume_lr'],
classification=cnf['classification'],
gpu_memory_fraction=args['gpu_memory_fraction'],
num_classes=args['num_classes'],
is_summary=args['is_summary'],
loss_type=args['loss_type'],
weighted=args['weighted'],
log_file_name=args['log_file_name'],
verbosity=args['verbose'],
is_early_stop=cnf.get('is_early_stop', True))
_early_stop, _loss = learner.fit(data_set,
weights_from=weights_from,
start_epoch=args['start_epoch'],
weights_dir=args['weights_dir'],
summary_every=399)
return {'early_stop': _early_stop, 'loss': _loss}
def main(model, training_cnf, data_dir, parallel, start_epoch, weights_from,
weights_dir, resume_lr, gpu_memory_fraction, num_classes, is_summary,
loss_type, weighted, log_file_name):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir,
model_def.image_size[0],
mode=cnf.get('mode'),
multilabel=cnf.get('multilabel', False))
standardizer = cnf.get('standardizer', NoOpStandardizer())
cutout = cnf.get('cutout', None)
training_iter, validation_iter = create_training_iters(
cnf,
data_set,
standardizer,
model_def.crop_size,
start_epoch,
parallel=parallel,
cutout=cutout,
data_balancing=cnf.get('data_balancing', False))
learner = SupervisedLearner(model,
cnf,
training_iterator=training_iter,
validation_iterator=validation_iter,
resume_lr=resume_lr,
classification=cnf['classification'],
gpu_memory_fraction=gpu_memory_fraction,
num_classes=num_classes,
is_summary=is_summary,
loss_type=loss_type,
weighted=weighted,
log_file_name=log_file_name)
learner.fit(data_set,
weights_from,
start_epoch=start_epoch,
weights_dir=weights_dir,
summary_every=399)
def main(model, training_cnf, data_dir, start_epoch, resume_lr, weights_from,
weights_exclude_scopes, trainable_scopes, clean, visuals):
util.check_required_program_args([model, training_cnf, data_dir])
sys.path.insert(0, '.')
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
util.init_logging('train.log',
file_log_level=logging.INFO,
console_log_level=logging.INFO,
clean=clean)
if weights_from:
weights_from = str(weights_from)
data_set = DataSet(data_dir, model_def.image_size[0])
training_iter, validation_iter = create_training_iters(
cnf, data_set, model_def.crop_size, start_epoch,
cnf.get('iterator_type', 'parallel') == 'parallel')
try:
input_shape = (-1, model_def.crop_size[1], model_def.crop_size[0],
model_def.num_channels)
except AttributeError:
input_shape = (-1, model_def.crop_size[1], model_def.crop_size[0], 3)
trainer = SupervisedTrainerQ(model,
cnf,
input_shape,
trainable_scopes,
training_iter,
validation_iter,
classification=cnf['classification'])
trainer.fit(data_set,
weights_from,
weights_exclude_scopes,
start_epoch,
resume_lr,
verbose=1,
summary_every=cnf.get('summary_every', 10),
clean=clean,
visuals=visuals)
def predict(model, training_cnf, predict_dir, weights_from, dataset_name,
convert, image_size, sync, test_type):
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
weights_from = str(weights_from)
images = data.get_image_files(predict_dir)
standardizer = cnf.get('standardizer', None)
preprocessor = convert_preprocessor(image_size) if convert else None
prediction_iterator = create_prediction_iter(cnf, standardizer,
model_def.crop_size,
preprocessor, sync)
if test_type == 'quasi':
predictor = QuasiCropPredictor(model, cnf, weights_from,
prediction_iterator, 20)
predictions = predictor.predict(images)
if not os.path.exists(os.path.join(predict_dir, '..', 'results')):
os.mkdir(os.path.join(predict_dir, '..', 'results'))
if not os.path.exists(
os.path.join(predict_dir, '..', 'results', dataset_name)):
os.mkdir(os.path.join(predict_dir, '..', 'results', dataset_name))
names = data.get_names(images)
image_prediction_prob = np.column_stack([names, predictions])
headers = ['score%d' % (i + 1) for i in range(predictions.shape[1])]
title = np.array(['image'] + headers)
image_prediction_prob = np.vstack([title, image_prediction_prob])
labels_file_prob = os.path.abspath(
os.path.join(predict_dir, '..', 'results', dataset_name,
'predictions.csv'))
np.savetxt(labels_file_prob,
image_prediction_prob,
delimiter=",",
fmt="%s")
def predict(frozen_model, training_cnf, image_path, image_size, output_path, gpu_memory_fraction):
cnf = util.load_module(training_cnf).cnf
standardizer = cnf['standardizer']
graph = util.load_frozen_graph(frozen_model)
preprocessor = convert_preprocessor(448)
predictor = SegmentPredictor(graph, standardizer, preprocessor)
final_prediction_map = predictor.predict(image_path)
final_prediction_map = final_prediction_map.transpose(0, 2, 1).squeeze()
image = data.load_image(image_path, preprocessor=preprocessor)
img = image.transpose(2, 1, 0)
img = Image.fromarray(img.astype('uint8'), 'RGB')
img.save('/tmp/test.png')
image_filename = image_path.split('/')[-1]
plot_masks('/tmp/test.png', final_prediction_map, output_path)
"""
def save_graph(model, output_dir, output_model):
model_def = util.load_module(model)
model = model_def.model
try:
with tf.Graph().as_default():
sess = tf.Session()
end_points_predict = model(is_training=False, reuse=None)
inputs = end_points_predict['inputs']
predictions = end_points_predict['predictions']
init = tf.global_variables_initializer()
sess.run(init)
tf.train.write_graph(sess.graph_def, output_dir, output_model)
except Exception as e:
print(e.message)
def predict(frozen_model, training_cnf, predict_dir, image_size, output_path,
num_classes, gpu_memory_fraction):
cnf = util.load_module(training_cnf).cnf
standardizer = cnf['standardizer']
graph = util.load_frozen_graph(frozen_model)
preprocessor = convert_preprocessor(image_size)
predictor = SegmentPredictor(graph, standardizer, preprocessor)
# images = data.get_image_files(predict_dir)
image_names = [
filename.strip() for filename in os.listdir(predict_dir)
if filename.endswith('.jpg')
]
iou = IOU()
per_class_iou = iou.per_class_iou(predictor, predict_dir, image_size)
meaniou = iou.meaniou(predictor, predict_dir, image_size)
print(per_class_iou)
print('Mean IOU %5.5f' % meaniou)
def main(training_cnf, interactive, gpu_memory_fraction):
"""Train a word2vec model."""
cnf = util.load_module(training_cnf).cnf
gpu_options = tf.GPUOptions(
per_process_gpu_memory_fraction=gpu_memory_fraction)
with tf.Graph().as_default(), tf.Session(config=tf.ConfigProto(
gpu_options=gpu_options)) as sess:
with tf.device("/cpu:0"):
model = Word2Vec(cnf, sess)
model.read_analogies() # Read analogy questions
for _ in xrange(cnf.get('epochs_to_train')):
model.train() # Process one epoch
model.eval() # Eval analogies.
# Perform a final save.
model.saver.save(sess,
os.path.join(cnf.get('save_path'), "model.ckpt"),
global_step=model.global_step)
if interactive:
# E.g.,
# [0]: model.analogy(b'france', b'paris', b'russia')
# [1]: model.nearby([b'proton', b'elephant', b'maxwell'])
_start_shell(locals())
def try_config(args, cnf):
"""For trying out configurations.
Args:
args: command line arguments regarding training
cnf: training configuration sampled from hyperband search space
Returns:
a dictionary containing final loss value and early stop flag
"""
mnist = input_data.read_data_sets("MNIST_data/", one_hot=False)
width = 28
height = 28
train_images = mnist[0].images
train_labels = mnist[0].labels
validation_images = mnist[1].images
validation_labels = mnist[1].labels
data_set = DataSet(train_images, train_labels, validation_images,
validation_labels)
model_def = util.load_module(args['model'])
model = model_def.model
learner = SupervisedLearner(model,
cnf,
classification=cnf['classification'],
is_summary=False,
num_classes=10,
verbosity=args['verbose'],
is_early_stop=cnf.get('is_early_stop', True))
_early_stop, _loss = learner.fit(data_set,
weights_from=None,
start_epoch=1)
return {'early_stop': _early_stop, 'loss': _loss}
def test_model(model, input_shape, loss_type='softmax'):
print(input_shape)
input_shape = [c.strip() for c in input_shape.split(",")]
model_def = util.load_module(model)
model = model_def.model
inputs = tf.placeholder(tf.float32, input_shape)
end_points = model(inputs, True, None)
logits = end_points['logits']
labels = tf.placeholder(tf.int32, (None, logits.get_shape().as_list()[1]))
prediction = end_points['prediction']
if loss_type == 'sigmoid':
loss = tf.losses.sigmoid_cross_entropy(labels, logits)
else:
loss = tf.losses.softmax_cross_entropy(labels, logits)
optimizer = tf.train.AdamOptimizer()
train_op = optimizer.minimize(loss)
feed_dict = {
inputs: np.random.normal(size=inputs.get_shape().as_list()),
labels: np.random.randint(0, high=1, size=logits.get_shape().as_list())
}
mltest.test_suite(prediction,
train_op,
feed_dict=feed_dict,
output_range=(0, 1))
def predict(model, output_layer, training_cnf, predict_dir, weights_from, tag,
convert, image_size, sync, predict_type):
util.check_required_program_args(
[model, training_cnf, predict_dir, weights_from])
model_def = util.load_module(model)
model = model_def.model
cnf = util.load_module(training_cnf).cnf
weights_from = str(weights_from)
images = data.get_image_files(predict_dir)
preprocessor = convert_preprocessor(image_size) if convert else None
prediction_iterator = create_prediction_iter(cnf, model_def.crop_size,
preprocessor, sync)
if predict_type == 'quasi':
predictor = QuasiCropPredictor(model, cnf, weights_from,
prediction_iterator, 20, output_layer)
elif predict_type == '1_crop':
predictor = OneCropPredictor(model, cnf, weights_from,
prediction_iterator, output_layer)
elif predict_type == '10_crop':
predictor = TenCropPredictor(model, cnf, weights_from,
prediction_iterator,
model_def.crop_size[0],
model_def.image_size[0], output_layer)
else:
raise ValueError('Unknown predict_type: %s' % predict_type)
predictions = predictor.predict(images)
prediction_results_dir = os.path.abspath(
os.path.join(predict_dir, '..', 'predictions', tag))
if not os.path.exists(prediction_results_dir):
os.makedirs(prediction_results_dir)
if output_layer == 'predictions':
names = data.get_names(images)
image_prediction_probs = np.column_stack([names, predictions])
headers = ['score%d' % (i + 1) for i in range(predictions.shape[1])]
title = np.array(['image'] + headers)
image_prediction_probs = np.vstack([title, image_prediction_probs])
prediction_probs_file = os.path.join(prediction_results_dir,
'predictions.csv')
np.savetxt(prediction_probs_file,
image_prediction_probs,
delimiter=",",
fmt="%s")
print('Predictions saved to: %s' % prediction_probs_file)
if cnf['classification']:
class_predictions = np.argmax(predictions, axis=1)
image_class_predictions = np.column_stack(
[names, class_predictions])
title = np.array(['image', 'label'])
image_class_predictions = np.vstack(
[title, image_class_predictions])
prediction_class_file = os.path.join(prediction_results_dir,
'predictions_class.csv')
np.savetxt(prediction_class_file,
image_class_predictions,
delimiter=",",
fmt="%s")
print('Class predictions saved to: %s' % prediction_class_file)
else:
# feature extraction
features_file = os.path.join(prediction_results_dir, 'features.npy')
np.save(features_file, predictions)
print('Features from layer: %s saved to: %s' %
(output_layer, features_file))
def predict(model, training_cnf, predict_dir, weights_from, dataset_name,
convert, image_size, sync, predict_type):
images = data.get_image_files(predict_dir)
# Form now, hard coded models, cnfs, and weights
# Need to take these from program inputs or an ensembling config file
print('Creating predictor 1')
weights_from1 = 'weights.sa/model-epoch-97.ckpt'
model1 = 'examples/mnist_model_sa.py'
training_cnf1 = 'examples/mnist_cnf.py'
model_def1 = util.load_module(model1)
model1 = model_def1.model
cnf1 = util.load_module(training_cnf1).cnf
standardizer = cnf1.get('standardizer', NoOpStandardizer())
preprocessor = convert_preprocessor(
model_def1.image_size[0]) if convert else None
prediction_iterator1 = create_prediction_iter(cnf1, standardizer,
model_def1.crop_size,
preprocessor, sync)
predictor1 = QuasiCropPredictor(model1, cnf1, weights_from1,
prediction_iterator1, 20)
# predictor1 = OneCropPredictor(model1, cnf1, weights_from1, prediction_iterator1)
print('Creating predictor 2')
weights_from2 = 'weights.rv/model-epoch-31.ckpt'
model2 = 'examples/mnist_model.py'
training_cnf2 = 'examples/mnist_cnf.py'
model_def2 = util.load_module(model2)
model2 = model_def2.model
cnf2 = util.load_module(training_cnf2).cnf
standardizer = cnf2.get('standardizer', NoOpStandardizer())
preprocessor = convert_preprocessor(
model_def2.image_size[0]) if convert else None
prediction_iterator2 = create_prediction_iter(cnf2, standardizer,
model_def2.crop_size,
preprocessor, sync)
predictor2 = QuasiCropPredictor(model2, cnf2, weights_from2,
prediction_iterator2, 20)
# predictor2 = OneCropPredictor(model2, cnf2, weights_from2, prediction_iterator2)
predictor = EnsemblePredictor([predictor1, predictor2])
predictions = predictor.predict(images)
if not os.path.exists(os.path.join(predict_dir, '..', 'results')):
os.mkdir(os.path.join(predict_dir, '..', 'results'))
if not os.path.exists(
os.path.join(predict_dir, '..', 'results', dataset_name)):
os.mkdir(os.path.join(predict_dir, '..', 'results', dataset_name))
names = data.get_names(images)
image_prediction_probs = np.column_stack([names, predictions])
headers = ['score%d' % (i + 1) for i in range(predictions.shape[1])]
title = np.array(['image'] + headers)
image_prediction_probs = np.vstack([title, image_prediction_probs])
prediction_probs_file = os.path.abspath(
os.path.join(predict_dir, '..', 'results', dataset_name,
'predictions.csv'))
np.savetxt(prediction_probs_file,
image_prediction_probs,
delimiter=",",
fmt="%s")
print('Predictions saved to: %s' % prediction_probs_file)
if cnf1['classification']:
class_predictions = np.argmax(predictions, axis=1)
image_class_predictions = np.column_stack([names, class_predictions])
title = np.array(['image', 'label'])
image_class_predictions = np.vstack([title, image_class_predictions])
prediction_class_file = os.path.abspath(
os.path.join(predict_dir, '..', 'results', dataset_name,
'predictions_class.csv'))
np.savetxt(prediction_class_file,
image_class_predictions,
delimiter=",",
fmt="%s")
print('Class predictions saved to: %s' % prediction_class_file)