parameter_dict = configFile.CONFIG_DICT[config_file_str]
parameter_dict[
'RESULTS_DIR'] = parameter_dict['RESULTS_DIR'] + '_' + model_type
use_gpu = torch.cuda.is_available() and parameter_dict['USE_GPU']
device = torch.device("cuda:0" if use_gpu else "cpu")
kwargs_testing = {}
kwargs_generator = {
'num_workers': 1,
'pin_memory': use_gpu,
'worker_init_fn': worker_init_fn
}
create_results_dir(parameter_dict['RESULTS_DIR'])
attach = True if parameter_dict['STARTING_EPOCH'] > 0 else False
experimentWriter = ExperimentWriter(join(parameter_dict['RESULTS_DIR'],
'experiment.txt'),
attach=attach)
###################################
########### DATA LOADER ###########
###################################
data_loader = DataLoader(parameter_dict)
sbj = data_loader.subject_list[0]
nslices = len(sbj.slice_list)
idx = np.random.choice(nslices)
data_loader.subject_list = sbj.slice_list[idx:idx + number_of_subjects]
data_loader.rid_list = [s.id for s in sbj.slice_list]
parser.add_argument("-p", help="Params", default=None, type=str)
arg = parser.parse_args(sys.argv[1:])
print('Getting parameters to train the model...')
params_string = arg.p
params = p.PARAMS_DICT[params_string].get_params()
filename = params[p.MODEL_NAME]
dir_path = join(params[p.OUTPUT_PATH],
'LR_' + str(params[p.LR]) + '_full_DA_shortcutTrue_allDB')
logs_filepath = join(dir_path, 'logs', filename + '.txt')
weights_filepath = join(dir_path, 'model_weights', filename + '.h5')
""" REDIRECT STDOUT TO FILE """
print('Output redirected to file... ')
print('Suggestion: Use tail command to see the output')
io.create_results_dir(dir_path=dir_path)
io.redirect_stdout_to_file(filepath=logs_filepath)
print('PARAMETERS')
print(params)
print('Learning rate exponential decay')
print('One subepoch')
print('double l1, l2')
""" ARCHITECTURE DEFINITION """
num_modalities = 2
model, output_shape = iSeg_models.get_model(
num_modalities=num_modalities,
segment_dimensions=tuple(params[p.INPUT_DIM]),
num_classes=params[p.N_CLASSES],
model_name=params[p.MODEL_NAME],
shortcut_input=params[p.SHORTCUT_INPUT],
from os.path import join
from os import listdir
import pdb
import shutil
from database import read_slice_info
from setup import DATA_DIR
from src.utils.io import create_results_dir
data_dir = join(DATA_DIR, 'dataset')
create_results_dir(data_dir, subdirs=['ihc','mri','nissl'])
shutil.copy('slice_info_ihc.csv', join(data_dir, 'ihc', 'slice_separation.csv'))
shutil.copy('slice_info_nissl.csv', join(data_dir, 'nissl', 'slice_separation.csv'))
slice_dict = read_slice_info(key='slice_number')
for stain in ['NISSL', 'IHC']:
files = listdir(join(DATA_DIR, stain.lower(), 'images_orig'))
for f in files:
slice_num = str(int(f.split('_')[1].split('.')[0]))
pdb.set_trace()
outf = slice_dict[stain][slice_num]['filename']
shutil.move(join(DATA_DIR, stain.lower(), 'images_orig', f),join(DATA_DIR, stain.lower(), 'images', outf))
shutil.move(join(DATA_DIR, stain.lower(), 'masks_orig', f),join(DATA_DIR, stain.lower(), 'masks', outf))
shutil.rmtree(join(DATA_DIR, stain.lower(), 'images_orig'))
shutil.rmtree(join(DATA_DIR, stain.lower(), 'masks_orig'))
INIT_BASE_DIR = join(DATA_DIR, 'downloads')
INIT_BASE_DIR_MRI = join(INIT_BASE_DIR, 'mri')
INIT_BASE_DIR_IHC = join(INIT_BASE_DIR, 'ihc')
INIT_BASE_DIR_NISSL = join(INIT_BASE_DIR, 'nissl')
INIT_BASE_DIR_LINEAL = join(INIT_BASE_DIR, 'linear')
MRI_ORIG = join(INIT_BASE_DIR_LINEAL, 'mri.orig.nii.gz')
MRI_LIN = join(INIT_BASE_DIR_LINEAL, 'mri.nii.gz')
HISTO_LIN = join(INIT_BASE_DIR_LINEAL, 'stack_nissl.lin.tree.nii.gz')
BASE_DIR = join(DATA_DIR, 'dataset')
BASE_DIR_MRI = join(BASE_DIR, 'mri')
BASE_DIR_IHC = join(BASE_DIR, 'ihc')
BASE_DIR_NISSL = join(BASE_DIR, 'nissl')
create_results_dir(BASE_DIR_MRI, subdirs=['images', 'masks'])
create_results_dir(
BASE_DIR_IHC,
subdirs=['images_resize', 'masks_resize', 'images', 'masks', 'affine'])
create_results_dir(
BASE_DIR_NISSL,
subdirs=['images_resize', 'masks_resize', 'images', 'masks', 'affine'])
#######################
###### Parameters #####
#######################
MRI_RES = 0.5
HISTO_RES = 0.032
HISTO_THICKNESS = 0.05
OUTPUT_RES = 0.25