train_input_names, train_output_names, val_input_names, val_output_names, test_input_names, test_output_names = prepare_data(
)
class_names_list = helpers.get_class_list(
os.path.join(args.dataset, "class_list.txt"))
class_names_string = ""
for class_name in class_names_list:
if not class_name == class_names_list[-1]:
class_names_string = class_names_string + class_name + ", "
else:
class_names_string = class_names_string + class_name
num_classes = len(class_names_list)
if args.balanced_weight:
b_weight = utils.median_frequency_balancing(
args.dataset + "/train_labels/", num_classes)
print("Preparing the model ...")
input = tf.placeholder(tf.float32, shape=[None, None, None, 3])
output = tf.placeholder(tf.float32, shape=[None, None, None, num_classes])
with tf.device('/gpu:' + str(args.gpu)):
input = tf.placeholder(tf.float32, shape=[None, None, None, 3])
output = tf.placeholder(tf.float32, shape=[None, None, None, num_classes])
if args.model == "FC-DenseNet56" or args.model == "FC-DenseNet67" or args.model == "FC-DenseNet103" or args.model == "FC-DenseNet158" or args.model == "FC-DenseNet232":
network = build_fc_densenet(input,
preset_model=args.model,
num_classes=num_classes,
dropout_p=0.5)
elif args.model == "RefineNet-Res101" or args.model == "RefineNet-Res152":
import cv2 as cv
import utils
map_data_dir = 'CamVid/train_labels/'
map_weihgt_dir = 'CamVid/train_labels_weights/'
if os.path.exists(map_weihgt_dir):
shutil.rmtree(map_weihgt_dir)
os.makedirs(map_weihgt_dir)
# get filenames
map_fns = np.asarray(sorted(glob.glob('%s/*.png*' % map_data_dir)))
n_all_files = len(map_fns)
print 'n_all_files:', n_all_files
balance_weight, cls_freq = utils.median_frequency_balancing(
'CamVid/train_labels/')
for map_fn in map_fns:
print map_weihgt_dir + os.path.basename(map_fn).split('.')[0] + '.png'
map_im = cv.imread(map_fn, cv.IMREAD_GRAYSCALE)
# _, binary = cv.threshold(11-map_im, 0, 255, cv.THRESH_BINARY)
# canny = cv.Canny(binary, 60, 0, apertureSize = 3)
# _,contours,_ = cv.findContours(canny, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
# blank = np.ones(map_im.shape)
# cv.drawContours(blank, contours, -1, 5, 1)
# pixel_weight = blank+2*(binary/255)
pixel_weight = np.empty(map_im.shape, dtype='float32')
for c in range(len(cls_freq)):
if c != 12:
pixel_weight[map_im == c] = max(2, 1 / cls_freq[c])
else:
"FC-DenseNet56", "FC-DenseNet67", "FC-DenseNet103", "FC-DenseNet158",
"Encoder-Decoder", "Encoder-Decoder-Skip", "RefineNet-Res101",
"RefineNet-Res152", "HF-FCN", "custom"
]
if args.model not in AVAILABLE_MODELS:
print("Error: given model is not available. Try these:")
print(AVAILABLE_MODELS)
print("Now exiting ...")
sys.exit()
# Load the data
print("Loading the data ...")
train_input_names, train_output_names, val_input_names, val_output_names, test_input_names, test_output_names = prepare_data(
)
if args.balanced_weight:
balanced_weight = utils.median_frequency_balancing(args.dataset +
"/train_labels/")
class_names_list = helpers.get_class_list(
os.path.join(args.dataset, "class_list.txt"))
class_names_string = ""
for class_name in class_names_list:
if not class_name == class_names_list[-1]:
class_names_string = class_names_string + class_name + ", "
else:
class_names_string = class_names_string + class_name
# num_classes = len(class_names_list)
num_classes = 1
print("Preparing the model ...")
input = tf.placeholder(tf.float32, shape=[None, None, None, 3])
print("Loading the data ...")
train_input_names, train_output_names, val_input_names, val_output_names, test_input_names, test_output_names = prepare_data(
)
class_names_list = helpers.get_class_list(
os.path.join(args.dataset, "class_list.txt"))
class_names_string = ""
for class_name in class_names_list:
if not class_name == class_names_list[-1]:
class_names_string = class_names_string + class_name + ", "
else:
class_names_string = class_names_string + class_name
num_classes = len(class_names_list)
if args.balanced_weight:
b_weight = utils.median_frequency_balancing(args.dataset + "/../gts_numpy",
num_classes)
print('b_weight: ', b_weight)
print("Preparing the model ...")
input = tf.placeholder(tf.float32, shape=[None, None, None, 3])
output = tf.placeholder(tf.float32, shape=[None, None, None, num_classes])
with tf.device('/gpu:' + str(args.gpu)):
input = tf.placeholder(tf.float32, shape=[None, None, None, 3])
output = tf.placeholder(tf.float32, shape=[None, None, None, num_classes])
if args.model == "FC-DenseNet56" or args.model == "FC-DenseNet67" or args.model == "FC-DenseNet103" or args.model == "FC-DenseNet158" or args.model == "FC-DenseNet232":
network = build_fc_densenet(input,
preset_model=args.model,
num_classes=num_classes)
elif args.model == "RefineNet-Res101" or args.model == "RefineNet-Res152":