val_dst = LeatherData(path_mask=path_val,
path_img=path_val,
list_of_filenames=file_names_val,
transform=transform_function,
color_dict=color_dict,
target_dict=target_dict)
train_images = []
if data_set == 'train':
for i in range(len(train_dst)):
train_images.append(train_dst.__getitem__(i))
elif data_set == 'val':
for i in range(len(val_dst)):
train_images.append(val_dst.__getitem__(i))
labels = [
'02', 'Abassamento', 'Abbassamento', 'Area Punture insetti', 'Area aperta',
'Area vene', 'Buco', 'Cicatrice', 'Cicatrice aperta', 'Contaminazione',
'Crease', 'Difetto di lavorazione', 'Dirt', 'Fianco', 'Fiore marcio',
'Insect bite', 'Marchio', 'Microcut', 'Piega', 'Pinza', 'Pinze', 'Poro',
"Puntura d'insetto", 'Puntura insetto', 'Ruga', 'Rughe', 'Scopertura',
'Scratch', 'Smagliatura', 'Soffiatura', 'Struttura', 'Taglio', 'Vena',
'Vene', 'Verruca', 'Wart', 'Zona aperta', 'verruca'
]
metrics = [StreamSegMetrics(2), StreamSegMetrics(2), StreamSegMetrics(2)]
false_positives = 0
true_negatives = [0, 0]
errors = np.array([[0, 0], [0, 0]])
num_workers=0)
val_loader = data.DataLoader(val_dst,
batch_size=val_batch_size,
shuffle=False,
num_workers=0)
# Load dataloader for unlabelled data:
trainloader_nl, _ = get_data_loaders_unlabelled(binary,
path_original_data,
path_meta_data,
dataset_path_ul,
batch_size,
size=SIZE)
train_img = []
for i in range(2):
train_img.append(train_dst.__getitem__(i))
print("Train set: %d, Val set: %d" % (len(train_dst), len(val_dst)))
if model_name == '':
model_name = 'DeepLab'
#model_name =
if optimizer == '':
optimizer = 'Adam'
if exp_descrip == '':
exp_descrip = 'no_decrip'
if train_scope == '':
train_scope = True
#training(n_classes=1, model="MobileNet", load_models=False, model_path=path_model,train_loader=train_loader, val_loader=val_loader, train_dst=train_dst, val_dst=val_dst,save_path=save_path, lr=lr, train_images=train_img, color_dict=color_dict, target_dict=target_dict,annotations_dict=annotations_dict,exp_description='tick')
training(n_classes=1,
model=model_name,