learning_rate = opt.lr #0.0001
num_channels = 15
piece_map = {}
piece_map['1_1_1'] = [0, 96, 0, 96, 0, 60]
train_source_dir = os.path.join(train_data_dir, 'source')
train_target_dir = os.path.join(train_data_dir, 'target')
test_source_dir = os.path.join(test_data_dir, 'source')
working_dir = os.path.join(working_root_dir, piece)
# define paths
out = os.path.join(working_dir, 'finetune_out')
mkdir(out)
train_source_subs, train_source_files = subl.get_sub_list(train_source_dir)
train_target_subs, train_target_files = subl.get_sub_list(train_target_dir)
train_dict = {}
train_dict['source_subs'] = train_source_subs
train_dict['source_files'] = train_source_files
train_dict['target_subs'] = train_target_subs
train_dict['target_files'] = train_target_files
test_source_subs, test_source_files = subl.get_sub_list(test_source_dir)
test_dict = {}
test_dict['source_subs'] = test_source_subs
test_dict['source_files'] = test_source_files
# load image
train_set = torchsrc.imgloaders.pytorch_loader_allpiece(
train_dict, num_channels=num_channels, piece=piece, piece_map=piece_map)
# train_dict = {}
# train_dict['img_subs'] = train_img_subs
# train_dict['img_files'] = train_img_files
# train_dict['seg_subs'] = train_seg_subs
# train_dict['seg_files'] = train_seg_files
else:
out = os.path.join(working_dir, 'test_out')
mkdir(out)
# train_img_subs, train_img_files, train_seg_subs, train_seg_files = subl.get_sub_from_txt(train_list_file, trainnii_list_file, label_list_file, labelnii_list_file)
#train_dict = {}
#train_dict['img_subs'] = train_img_subs
#train_dict['img_files'] = train_img_files
#train_dict['seg_subs'] = train_seg_subs
#train_dict['seg_files'] = train_seg_files
test_img_subs, test_img_files = subl.get_sub_list(test_img_dir)
test_dict = {}
test_dict['img_subs'] = test_img_subs
test_dict['img_files'] = test_img_files
# load image
#train_set = torchsrc.imgloaders.pytorch_loader_allpiece(train_dict, num_labels=lmk_num, piece=piece, piece_map=piece_map)
#train_loader = torch.utils.data.DataLoader(train_set, batch_size=batch_size, shuffle=True, num_workers=1)
test_set = torchsrc.imgloaders.pytorch_loader_allpiece(test_dict,
num_labels=lmk_num,
piece=piece,
piece_map=piece_map)
test_loader = torch.utils.data.DataLoader(test_set,
batch_size=batch_size,
shuffle=False,
num_workers=1)
piece_map['2_2_1'] = [38, 134, 46, 174, 0, 88]
piece_map['2_2_3'] = [38, 134, 46, 174, 68, 156]
# define paths
train_list_file = '/share4/huoy1/Deep_5000_Brain/sublist/sublist_mni.txt'
working_dir = os.path.join('/share4/huoy1/Deep_5000_Brain/working_dir/', piece)
test_img_dir = '/share4/huoy1/Deep_5000_Brain/testing/mni/T1'
finetune_img_dir = '/share4/huoy1/Deep_5000_Brain/finetune_training/aladin-reg-images-normalized'
finetune_seg_dir = '/share4/huoy1/Deep_5000_Brain/finetune_training/aladin-reg-labels/'
# make img list
if finetune == True:
out = os.path.join(working_dir, 'finetune_out')
mkdir(out)
train_img_subs, train_img_files = subl.get_sub_list(finetune_img_dir)
train_seg_subs, train_seg_files = subl.get_sub_list(finetune_seg_dir)
train_dict = {}
train_dict['img_subs'] = train_img_subs
train_dict['img_files'] = train_img_files
train_dict['seg_subs'] = train_seg_subs
train_dict['seg_files'] = train_seg_files
else:
out = os.path.join(working_dir, 'test_out')
mkdir(out)
train_img_subs, train_img_files, train_seg_subs, train_seg_files = subl.get_sub_from_txt(
train_list_file)
train_dict = {}
train_dict['img_subs'] = train_img_subs
train_dict['img_files'] = train_img_files
train_dict['seg_subs'] = train_seg_subs