def evaluate(eval_params, net_params, data_params, common_params, train_params):
eval_model_path = eval_params['eval_model_path']
num_classes = net_params['num_class']
labels = data_params['labels']
data_dir = eval_params['data_dir']
label_dir = eval_params['label_dir']
volumes_txt_file = eval_params['volumes_txt_file']
remap_config = eval_params['remap_config']
device = common_params['device']
log_dir = common_params['log_dir']
exp_dir = common_params['exp_dir']
exp_name = train_params['exp_name']
save_predictions_dir = eval_params['save_predictions_dir']
prediction_path = os.path.join(exp_dir, exp_name, save_predictions_dir)
orientation = eval_params['orientation']
data_id = eval_params['data_id']
logWriter = LogWriter(num_classes, log_dir, exp_name, labels=labels)
avg_dice_score, class_dist = eu.evaluate_dice_score(eval_model_path,
num_classes,
data_dir,
label_dir,
volumes_txt_file,
remap_config,
orientation,
prediction_path,
data_id,
device,
logWriter)
logWriter.close()
def __init__(self,
model,
exp_name,
device,
num_class,
optim=torch.optim.SGD,
optim_args={},
loss_func=losses.DiceLoss(),
model_name='OneShotSegmentor',
labels=None,
num_epochs=10,
log_nth=5,
lr_scheduler_step_size=5,
lr_scheduler_gamma=0.5,
use_last_checkpoint=True,
exp_dir='experiments',
log_dir='logs'):
self.device = device
self.model = model
self.model_name = model_name
self.labels = labels
self.num_epochs = num_epochs
if torch.cuda.is_available():
self.loss_func = loss_func.cuda(device)
else:
self.loss_func = loss_func
self.optim_c = optim(
[{'params': model.conditioner.parameters(), 'lr': 1e-3, 'momentum': 0.99, 'weight_decay': 0.0001}
], **optim_args)
self.optim_s = optim(
[{'params': model.segmentor.parameters(), 'lr': 1e-3, 'momentum': 0.99, 'weight_decay': 0.0001}
], **optim_args)
self.scheduler_s = lr_scheduler.StepLR(self.optim_s, step_size=10,
gamma=0.1)
self.scheduler_c = lr_scheduler.StepLR(self.optim_c, step_size=10,
gamma=0.001)
exp_dir_path = os.path.join(exp_dir, exp_name)
common_utils.create_if_not(exp_dir_path)
common_utils.create_if_not(os.path.join(exp_dir_path, CHECKPOINT_DIR))
self.exp_dir_path = exp_dir_path
self.log_nth = log_nth
self.logWriter = LogWriter(
num_class, log_dir, exp_name, use_last_checkpoint, labels)
self.use_last_checkpoint = use_last_checkpoint
self.start_epoch = 1
self.start_iteration = 1
self.best_ds_mean = 0
self.best_ds_mean_epoch = 0
if use_last_checkpoint:
self.load_checkpoint()
def __init__(self,
model,
exp_name,
device,
num_class,
optim=torch.optim.SGD,
optim_args={},
loss_func=additional_losses.CombinedLoss(),
model_name='quicknat',
labels=None,
num_epochs=10,
log_nth=5,
lr_scheduler_step_size=5,
lr_scheduler_gamma=0.5,
use_last_checkpoint=True,
exp_dir='experiments',
log_dir='logs',
arch_file_path=None):
self.device = device
self.model = model
# self.swa_model = torch.optim.swa_utils.AveragedModel(self.model)
self.model_name = model_name
self.labels = labels
self.num_epochs = num_epochs
if torch.cuda.is_available():
self.loss_func = loss_func.cuda(device)
else:
self.loss_func = loss_func
self.optim = optim(model.parameters(), **optim_args)
# self.scheduler = lr_scheduler.StepLR(self.optim, step_size=lr_scheduler_step_size,
# gamma=lr_scheduler_gamma)
self.scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(self.optim, T_max=100)
# self.swa_start = -1 #int(np.round(self.num_epochs*0.75))
# print(self.swa_start)
# self.swa_scheduler = torch.optim.swa_utils.SWALR(self.optim, swa_lr=0.05)
exp_dir_path = os.path.join(exp_dir, exp_name)
common_utils.create_if_not(exp_dir_path)
common_utils.create_if_not(os.path.join(exp_dir_path, CHECKPOINT_DIR))
self.exp_dir_path = exp_dir_path
self.save_architectural_files(arch_file_path)
self.log_nth = log_nth
self.logWriter = LogWriter(num_class, log_dir, exp_name, use_last_checkpoint, labels)
# self.wandb = wandb
self.use_last_checkpoint = use_last_checkpoint
self.start_epoch = 1
self.start_iteration = 1
self.best_ds_mean = 0
self.best_ds_mean_epoch = 0
if use_last_checkpoint:
self.load_checkpoint()
print(self.best_ds_mean, self.best_ds_mean_epoch, self.start_epoch)
def evaluate(eval_params, net_params, data_params, common_params, train_params):
eval_model_path = eval_params['eval_model_path']
num_classes = net_params['num_class']
labels = data_params['labels']
data_dir = eval_params['data_dir']
label_dir = eval_params['label_dir']
volumes_txt_file = eval_params['volumes_txt_file']
device = common_params['device']
log_dir = common_params['log_dir']
exp_dir = common_params['exp_dir']
exp_name = train_params['exp_name']
save_predictions_dir = eval_params['save_predictions_dir']
prediction_path = os.path.join(exp_dir, exp_name, save_predictions_dir)
orientation = eval_params['orientation']
multi_channel = data_params['use_3channel']
use_2channel = data_params['use_2channel']
thick_channel = data_params['thick_channel']
logWriter = LogWriter(num_classes, log_dir, exp_name, labels=labels)
avg_dice_score, class_dist = evaluate_dice_score(eval_model_path,
num_classes,
data_dir,
label_dir,
volumes_txt_file,
orientation,
prediction_path,
device,
logWriter,
multi_channel=multi_channel,
use_2channel=use_2channel,
thick_ch=thick_channel)
logWriter.close()
def __init__(self,
model,
exp_name,
device,
num_class,
optim=torch.optim.SGD,
optim_args={},
loss_func=losses.CombinedLoss(),
model_name='segmentor',
labels=None,
num_epochs=10,
log_nth=5,
lr_scheduler_step_size=5,
lr_scheduler_gamma=0.5,
use_last_checkpoint=True,
exp_dir='experiments',
log_dir='logs'):
self.device = device
self.model = model
self.model_name = model_name
self.labels = labels
self.num_epochs = num_epochs
if torch.cuda.is_available():
self.loss_func = loss_func.cuda(device)
else:
self.loss_func = loss_func
self.optim = optim(model.parameters(), **optim_args)
self.scheduler = lr_scheduler.StepLR(self.optim,
step_size=lr_scheduler_step_size,
gamma=lr_scheduler_gamma)
exp_dir_path = os.path.join(exp_dir, exp_name)
common_utils.create_if_not(exp_dir_path)
common_utils.create_if_not(os.path.join(exp_dir_path, CHECKPOINT_DIR))
self.exp_dir_path = exp_dir_path
self.log_nth = log_nth
self.logWriter = LogWriter(num_class, log_dir, exp_name,
use_last_checkpoint, labels)
self.use_last_checkpoint = use_last_checkpoint
self.start_epoch = 1
self.start_iteration = 1
self.best_ds_mean = 0
self.best_ds_mean_epoch = 0
if use_last_checkpoint:
self.load_checkpoint()
def evaluate(eval_params, net_params, data_params, common_params,
train_params):
eval_model_path = eval_params['eval_model_path']
num_classes = net_params['num_class']
labels = data_params['labels']
data_dir = eval_params['data_dir']
query_txt_file = eval_params['query_txt_file']
support_txt_file = eval_params['support_txt_file']
remap_config = eval_params['remap_config']
device = common_params['device']
log_dir = common_params['log_dir']
exp_dir = common_params['exp_dir']
exp_name = train_params['exp_name']
save_predictions_dir = eval_params['save_predictions_dir']
orientation = eval_params['orientation']
logWriter = LogWriter(num_classes, log_dir, exp_name, labels=labels)
folds = ['fold1', 'fold2', 'fold3', 'fold4']
for fold in folds:
prediction_path = os.path.join(exp_dir, exp_name)
prediction_path = prediction_path + "_" + fold
prediction_path = os.path.join(prediction_path, save_predictions_dir)
eval_model_path = os.path.join(
'saved_models', eval_model_path + '_' + fold + '.pth.tar')
query_labels = get_lab_list('val', fold)
num_classes = len(fold)
avg_dice_score = eu.evaluate_dice_score(eval_model_path,
num_classes,
query_labels,
data_dir,
query_txt_file,
support_txt_file,
remap_config,
orientation,
prediction_path,
device,
logWriter,
fold=fold)
logWriter.log(avg_dice_score)
logWriter.close()
def evaluate(eval_params, net_params, data_params, common_params,
train_params):
eval_model_path = eval_params['eval_model_path']
num_classes = net_params['num_class']
labels = data_params['labels']
data_dir = eval_params['data_dir']
label_dir = eval_params['label_dir']
query_txt_file = eval_params['query_txt_file']
support_txt_file = eval_params['support_txt_file']
remap_config = eval_params['remap_config']
device = common_params['device']
log_dir = common_params['log_dir']
exp_dir = common_params['exp_dir']
exp_name = train_params['exp_name']
save_predictions_dir = eval_params['save_predictions_dir']
prediction_path = os.path.join(exp_dir, exp_name, save_predictions_dir)
orientation = eval_params['orientation']
logWriter = LogWriter(num_classes, log_dir, exp_name, labels=labels)
# model_name = 'model6_Dice_L2_loss_target_fold1.pth.tar'
folds = ['fold4']
# eval_model_path1 = "saved_models/sne_position_all_type_spatial_fold1.pth.tar"
eval_model_path1 = "finetuned_segmentor"
eval_model_path2 = "saved_models/model6_coronal_wholebody_condch16_e4Skip_inter_e3e4bnd4d3_ch_e1e2d1d2_noSseSeg_DiceLoss_lowrate_fold2.pth.tar"
# eval_model_path3 = "saved_models/model6_sagittal_fold1.pth.tar"
orientaion1 = 'AXI'
orientaion2 = 'COR'
for fold in folds:
eval_model_path = os.path.join(
'saved_models', eval_model_path1 + '_' + fold + '.pth.tar')
query_labels = _get_lab_list('val', fold)
num_classes = len(fold)
avg_dice_score = eu.evaluate_dice_score(eval_model_path,
num_classes,
query_labels,
data_dir,
query_txt_file,
support_txt_file,
remap_config,
orientaion1,
prediction_path,
device,
logWriter,
fold=fold)
# avg_dice_score = eu.evaluate_dice_score_3view(eval_model_path1,
# eval_model_path2,
# eval_model_path3,
# num_classes,
# query_labels,
# data_dir,
# query_txt_file,
# support_txt_file,
# remap_config,
# orientaion1,
# prediction_path,
# device,
# logWriter, fold=fold)
# avg_dice_score = eu.evaluate_dice_score_2view(eval_model_path1,
# eval_model_path2,
# num_classes,
# query_labels,
# data_dir,
# query_txt_file,
# support_txt_file,
# remap_config,
# orientaion1,
# prediction_path,
# device,
# logWriter, fold=fold)
logWriter.log(avg_dice_score)
logWriter.close()