def __init__(self, image_dir, label_dir, params):
train_filenames, train_label = get_filenames_and_labels(
image_dir, label_dir, 'train')
nones = [None] * len(train_filenames)
train_samples = list(zip(train_filenames, nones, train_label))
val_filenames, val_label = get_filenames_and_labels(
image_dir, label_dir, 'dev')
nones = [None] * len(val_filenames)
val_samples = list(zip(val_filenames, nones, val_label))
self.preset = preset = get_preset_by_name('ssdmobilenet160')
self.num_classes = 2
self.train_tfs = build_train_transforms(self.preset, self.num_classes)
self.val_tfs = build_val_transforms(self.preset, self.num_classes)
self.train_generator = self.__build_generator(train_samples,
self.train_tfs)
self.val_generator = self.__build_generator(val_samples, self.val_tfs)
self.num_train = len(train_samples)
params.train_size = self.num_train
self.num_val = len(val_samples)
params.eval_size = self.num_val
self.train_samples = list(zip(train_filenames, train_label))
self.val_samples = list(zip(val_filenames, val_label))
self.params = params
assert not overwritting, "Weights found in model_dir, aborting to avoid overwrite"
# Set the logger
set_logger(os.path.join(args.model_dir, 'train.log'))
# Create the input data pipeline
logging.info("Creating the datasets...")
data_dir = args.data_dir
model_dir = args.model_dir
image_dir = os.path.join(data_dir, 'Images')
label_dir = os.path.join(data_dir, 'Labels')
# Create the two iterators over the two datasets
train_inputs = input_fn(True, image_dir, label_dir, params)
eval_inputs = input_fn(False, image_dir, label_dir, params)
# Define the model
logging.info("Creating the model...")
preset = get_preset_by_name('ssdmobilenet160')
train_model_specs = model_fn('train', train_inputs, preset, params)
eval_model_specs = model_fn('eval',
eval_inputs,
preset,
params,
reuse=True)
# Train the model
logging.info("Starting training for {} epoch(s)".format(params.num_epochs))
train_and_evaluate(train_model_specs, eval_model_specs, model_dir, params,
args.restore_from)