Skip to content

GitLab

Commit 261f6b5a authored by Vaanathi Sundaresan's avatar Vaanathi Sundaresan
Browse files

command line call script

parent bc151fff
Hide whitespace changes
Inline Side-by-side
truenet_tumseg/scripts/truenet_tumseg 0 → 100644
View file @ 261f6b5a
#!/usr/bin/env fslpython
import argparse
import sys
import pkg_resources
import codecs
from truenet_tumseg.truenet_tumorseg import (truenet_tumseg_commands, truenet_tumseg_help_messages)
#=========================================================================================
# FSL TRUE_NET_TUMOR_SEGMENTATION
# Vaanathi Sundaresan
# 01-05-2021, Oxford
#=========================================================================================
if __name__ == "__main__":
desc_msgs = truenet_tumseg_help_messages.desc_descs()
epilog_msgs = truenet_tumseg_help_messages.epilog_descs()
parser = argparse.ArgumentParser(prog='truenet_tumseg', formatter_class=argparse.RawDescriptionHelpFormatter,
description=desc_msgs['mainparser'], epilog=epilog_msgs['mainparser'])
subparsers = parser.add_subparsers(dest="command")
parser_train = subparsers.add_parser('train', formatter_class=argparse.RawDescriptionHelpFormatter,
description=desc_msgs['train'], epilog=epilog_msgs['subparsers'])
requiredNamedtrain = parser_train.add_argument_group('Required named arguments')
requiredNamedtrain.add_argument('-i', '--inp_dir', type=str, required=True, help='Input directory containing training images')
requiredNamedtrain.add_argument('-l', '--label_dir', type=str, required=True, help='Directory containing lesion manual masks')
requiredNamedtrain.add_argument('-m', '--model_dir', type=str, required=True, help='Directory for saving model weights')
optionalNamedtrain = parser_train.add_argument_group('Optional named arguments')
optionalNamedtrain.add_argument('-modality', '--select_modality', nargs='+', type=int, required=False, default=[1, 1, 1, 1, 0], help='Input modalities available to train; format={FLAIR, T1, T1ce, T2, Others} (default=1 1 1 1 0)')
optionalNamedtrain.add_argument('-tr_prop', '--train_prop', type = float, required=False, default=0.8, help='Proportion of data used for training (default = 0.8)')
optionalNamedtrain.add_argument('-bfactor', '--batch_factor', type = int, required=False, default=10, help='No. of subjects considered for each mini-epoch (default = 10)')
optionalNamedtrain.add_argument('-nclass', '--num_classes', type = int, required=False, default=2,
help='No of classes to consider in the target labels; any additional class will be considered part of background (default=2)')
optionalNamedtrain.add_argument('-plane', '--acq_plane', type = str, required=False, default='all', help='Options: axial, sagittal, coronal, tc, all (default = all)')
optionalNamedtrain.add_argument('-da', '--data_augmentation', type = bool, required=False, default=True, help='Applying data augmentation (default=True)')
optionalNamedtrain.add_argument('-af', '--aug_factor', type = int, required=False, default=2, help='Data inflation factor for augmentation (default=2)')
optionalNamedtrain.add_argument('-sv_resume', '--save_resume_training', type = bool, required=False, default=False, help='Whether to save and resume training in case of interruptions (default-False)')
optionalNamedtrain.add_argument('-ilr', '--init_learng_rate', type = float, required=False, default=0.001, help='Initial LR to use in scheduler (default=0.001)')
optionalNamedtrain.add_argument('-lrm', '--lr_sch_mlstone', nargs='+', type=int, required=False, default=10, help='Milestones for LR scheduler (default=10)')
optionalNamedtrain.add_argument('-gamma', '--lr_sch_gamma', type = float, required=False, default=0.1, help='LR reduction factor in the LR scheduler (default=0.1)')
optionalNamedtrain.add_argument('-opt', '--optimizer', type = str, required=False, default='adam', help='Optimizer used for training. Options: adam, sgd (default=adam)')
optionalNamedtrain.add_argument('-eps', '--epsilon', type = float, required=False, default=1e-4, help='Epsilon for adam optimiser (default=1e-4)')
optionalNamedtrain.add_argument('-mom', '--momentum', type = float, required=False, default=0.9, help='Momentum for sgd optimiser (default=0.9)')
optionalNamedtrain.add_argument('-bs', '--batch_size', type = int, required=False, default=8, help='Batch size (default=8)')
optionalNamedtrain.add_argument('-ep', '--num_epochs', type = int, required=False, default=60, help='Number of epochs (default=60)')
optionalNamedtrain.add_argument('-es', '--early_stop_val', type = int, required=False, default=20, help='No. of epochs to wait for progress (early stopping) (default=20)')
optionalNamedtrain.add_argument('-sv_mod', '--save_full_model', type = bool, required=False, default=False, help='Saving the whole model instead of weights alone (default=False)')
optionalNamedtrain.add_argument('-cp_type', '--cp_save_type', type = str, required=False, default='last', help='Checkpoint saving options: best, last, everyN (default=last)')
optionalNamedtrain.add_argument('-cp_n', '--cp_everyn_N', type = int, required=False, default=10, help='If -cp_type=everyN, the N value (default=10)')
optionalNamedtrain.add_argument('-v', '--verbose', type = bool, required=False, default=False, help='Display debug messages (default=False)')
parser_train.set_defaults(func=truenet_tumseg_commands.train)
parser_evaluate = subparsers.add_parser('evaluate', formatter_class=argparse.RawDescriptionHelpFormatter,
description=desc_msgs['evaluate'], epilog=epilog_msgs['subparsers'])
requiredNamedevaluate = parser_evaluate.add_argument_group('Required named arguments')
requiredNamedevaluate.add_argument('-i', '--inp_dir', type=str, required=True, help='Input directory containing test images')
requiredNamedevaluate.add_argument('-m', '--model_name', type=str, required=True, help='Model basename with absolute path (will not be considered if optional argument -p=True)')
requiredNamedevaluate.add_argument('-o', '--output_dir', type=str, required=True, help='Directory for saving predictions')
optionalNamedevaluate = parser_evaluate.add_argument_group('Optional named arguments')
optionalNamedevaluate.add_argument('-modality', '--select_modality', nargs='+', type=int, required=False, default=[1, 1, 1, 1, 0], help='Input modalities available to train; format={FLAIR, T1, T1ce, T2, Others} (default=1 1 1 1 0)')
optionalNamedevaluate.add_argument('-p', '--pretrained_model', type=bool, required=False, default=False, help='Whether to use a standard pre-trained model from MICCAI BraTS 2020 training data (default=False)')
optionalNamedevaluate.add_argument('-nclass', '--num_classes', type = int, required=False, default=2,
help='Number of classes in the labels used for training the model (for both pretrained models, -nclass=2) (default=2)')
optionalNamedevaluate.add_argument('-post', '--postprocessing', type=bool, required=False, default=True, help='Whether to perform post-processing for TC region (default=True)')
optionalNamedevaluate.add_argument('-int', '--intermediate', type=bool, required=False, default=False, help='Saving intermediate predictionss (individual planes) for each subject (default=False)')
optionalNamedevaluate.add_argument('-cp_type', '--cp_load_type', type = str, required=False, default='last', help='Checkpoint to be loaded. Options: best, last, specific (default = last)')
optionalNamedevaluate.add_argument('-cp_n', '--cp_everyn_N', type = int, required=False, default=None, help='If -cp_type=specific, the N value (default=10)')
optionalNamedevaluate.add_argument('-v', '--verbose', type = bool, required=False, default=False, help='Display debug messages (default=False)')
parser_evaluate.set_defaults(func=truenet_tumseg_commands.evaluate)
parser_finetune = subparsers.add_parser('fine_tune', formatter_class=argparse.RawDescriptionHelpFormatter,
description=desc_msgs['fine_tune'], epilog=epilog_msgs['subparsers'])
requiredNamedft = parser_finetune.add_argument_group('Required named arguments')
requiredNamedft.add_argument('-i', '--inp_dir', type=str, required=True, help='Input directory containing training images')
requiredNamedft.add_argument('-l', '--label_dir', type=str, required=True, help='Directory containing lesion manual masks')
requiredNamedft.add_argument('-m', '--model_name', type=str, required=True, help='Model basename with absolute path (will not be considered if optional argument -p=True)')
requiredNamedft.add_argument('-o', '--output_dir', type=str, required=True, help='Output directory for saving fine-tuned models/weights')
optionalNamedft = parser_finetune.add_argument_group('Optional named arguments')
optionalNamedft.add_argument('-modality', '--select_modality', nargs='+', type=int, required=False, default=[1, 1, 1, 1, 0], help='Input modalities available to train; format={FLAIR, T1, T1ce, T2, Others} (default=1 1 1 1 0)')
optionalNamedft.add_argument('-p', '--pretrained_model', type=bool, required=False, default=False, help='Whether to use a standard pre-trained model from MICCAI BraTS 2020 training data (default=False)')
optionalNamedft.add_argument('-cpld_type', '--cp_load_type', type = str, required=False, default='last', help='Checkpoint to be loaded. Options: best, last, specific (default=last')
optionalNamedft.add_argument('-cpld_n', '--cpload_everyn_N', type = int, required=False, default=10, help='If -cpld_type=specific, the N value (default=10)')
optionalNamedft.add_argument('-ftlayers', '--ft_layers', nargs='+', type=int, required=False, default=2, help='Layers to fine-tune starting from the decoder (default=1 2)')
optionalNamedft.add_argument('-tr_prop', '--train_prop', type = float, required=False, default=0.8, help='Proportion of data used for training (default = 0.8)')
optionalNamedft.add_argument('-bfactor', '--batch_factor', type = int, required=False, default=10, help='No. of subjects considered for each mini-epoch (default = 10)')
optionalNamedft.add_argument('-plane', '--acq_plane', type = str, required=False, default='all', help='The plane in which the model needs to be fine-tuned. Options: axial, sagittal, coronal, tc, all (default=all)')
optionalNamedft.add_argument('-nclass', '--num_classes', type = int, required=False, default=2,
help='No. of classes to consider in the target labels; any additional class will be considered part of background (default=2)')
optionalNamedft.add_argument('-da', '--data_augmentation', type = bool, required=False, default=True, help='Applying data augmentation (default=True)')
optionalNamedft.add_argument('-af', '--aug_factor', type = int, required=False, default=2, help='Data inflation factor for augmentation (default=2)')
optionalNamedft.add_argument('-sv_resume', '--save_resume_training', type = bool, required=False, default=False, help='Whether to save and resume training in case of interruptions (default-False)')
optionalNamedft.add_argument('-ilr', '--init_learng_rate', type = float, required=False, default=0.0001, help='Initial LR to use for fine-tuning in scheduler (default=0.0001)')
optionalNamedft.add_argument('-lrm', '--lr_sch_mlstone', nargs='+', type=int, required=False, default=10, help='Milestones for LR scheduler (default=10)')
optionalNamedft.add_argument('-gamma', '--lr_sch_gamma', type = float, required=False, default=0.1, help='LR reduction factor in the LR scheduler (default=0.1)')
optionalNamedft.add_argument('-opt', '--optimizer', type = str, required=False, default='adam', help='Optimizer used for training. Options:adam, sgd (default=adam)')
optionalNamedft.add_argument('-eps', '--epsilon', type = float, required=False, default=1e-4, help='Epsilon for adam optimiser (default=1e-4)')
optionalNamedft.add_argument('-mom', '--momentum', type = float, required=False, default=0.9, help='Momentum for sgd optimiser (default=0.9)')
optionalNamedft.add_argument('-bs', '--batch_size', type = int, required=False, default=8, help='Batch size (default=8)')
optionalNamedft.add_argument('-ep', '--num_epochs', type = int, required=False, default=60, help='Number of epochs (default=60)')
optionalNamedft.add_argument('-es', '--early_stop_val', type = int, required=False, default=20, help='No. of epochs to wait for progress (early stopping) (default=20)')
optionalNamedft.add_argument('-sv_mod', '--save_full_model', type = bool, required=False, default=False, help='Saving the whole model instead of weights alone (default=False)')
optionalNamedft.add_argument('-cp_type', '--cp_save_type', type = str, required=False, default='last', help='Checkpoint saving options: best, last, everyN (default=last)')
optionalNamedft.add_argument('-cp_n', '--cp_everyn_N', type = int, required=False, default=10, help='If -cp_type=everyN, the N value')
optionalNamedft.add_argument('-v', '--verbose', type = bool, required=False, default=False, help='Display debug messages (default=False)')
parser_finetune.set_defaults(func=truenet_tumseg_commands.fine_tune)
parser_cv = subparsers.add_parser('cross_validate', formatter_class=argparse.RawDescriptionHelpFormatter,
description=desc_msgs['cross_validate'], epilog=epilog_msgs['subparsers'])
requiredNamedcv = parser_cv.add_argument_group('Required named arguments')
requiredNamedcv.add_argument('-i', '--inp_dir', type=str, required=True, help='Input directory containing images')
requiredNamedcv.add_argument('-l', '--label_dir', type=str, required=True, help='Directory containing lesion manual masks')
requiredNamedcv.add_argument('-o', '--output_dir', type=str, required=True, help='Output directory for saving predictions (and models)')
optionalNamedcv = parser_cv.add_argument_group('Optional named arguments')
optionalNamedcv.add_argument('-modality', '--select_modality', nargs='+', type=int, required=False, default=[1, 1, 1, 1, 0], help='Input modalities available to train; format={FLAIR, T1, T1ce, T2, Others} (default=1 1 1 1 0)')
optionalNamedcv.add_argument('-fold', '--cv_fold', type = int, required=False, default=5, help='Number of folds for cross-validation (default = 5)')
optionalNamedcv.add_argument('-resume_fold', '--resume_from_fold', type = int, required=False, default=1, help='Resume cross-validation from the specified fold (default = 1)')
optionalNamedcv.add_argument('-post', '--postprocessing', type=bool, required=False, default=True, help='Whether to perform post-processing for TC region (default=True)')
optionalNamedcv.add_argument('-tr_prop', '--train_prop', type = float, required=False, default=0.8, help='Proportion of data used for training (default = 0.8)')
optionalNamedcv.add_argument('-bfactor', '--batch_factor', type = int, required=False, default=10, help='No. of subjects considered for each mini-epoch (default = 10)')
optionalNamedcv.add_argument('-nclass', '--num_classes', type = int, required=False, default=2,
help='No. of classes to consider in the target labels; any additional class will be considered part of background (default=2)')
optionalNamedcv.add_argument('-plane', '--acq_plane', type = str, required=False, default='all', help='The plane in which the model needs to be trained. Options: axial, sagittal, coronal, tc, all (default=all)')
optionalNamedcv.add_argument('-da', '--data_augmentation', type = bool, required=False, default=True, help='Applying data augmentation (default=True)')
optionalNamedcv.add_argument('-af', '--aug_factor', type = int, required=False, default=2, help='Data inflation factor for augmentation (default=2)')
optionalNamedcv.add_argument('-sv_resume', '--save_resume_training', type = bool, required=False, default=False, help='Whether to save and resume training in case of interruptions (default-False)')
optionalNamedcv.add_argument('-ilr', '--init_learng_rate', type = float, required=False, default=0.001, help='Initial LR to use in scheduler (default=0.001)')
optionalNamedcv.add_argument('-lrm', '--lr_sch_mlstone', nargs='+', type=int, required=False, default=10, help='Milestones for LR scheduler (default=10)')
optionalNamedcv.add_argument('-gamma', '--lr_sch_gamma', type = float, required=False, default=0.1, help='LR reduction factor in the LR scheduler (default=0.1)')
optionalNamedcv.add_argument('-opt', '--optimizer', type = str, required=False, default='adam', help='Optimizer used for training. Options: adam, sgd (default=adam)')
optionalNamedcv.add_argument('-eps', '--epsilon', type = float, required=False, default=1e-4, help='Epsilon for adam optimiser (default=1e-4)')
optionalNamedcv.add_argument('-mom', '--momentum', type = float, required=False, default=0.9, help='Momentum for sgd optimiser (default=0.9)')
optionalNamedcv.add_argument('-bs', '--batch_size', type = int, required=False, default=8, help='Batch size (default=8)')
optionalNamedcv.add_argument('-ep', '--num_epochs', type = int, required=False, default=60, help='Number of epochs (default=60)')
optionalNamedcv.add_argument('-es', '--early_stop_val', type = int, required=False, default=20, help='No. of epochs to wait for progress (early stopping) (default=20)')
optionalNamedcv.add_argument('-int', '--intermediate', type=bool, required=False, default=False, help='Saving intermediate prediction results for each subject (default=False)')
optionalNamedcv.add_argument('-sv', '--save_checkpoint', type = bool, required=False, default=False, help='Whether to save any checkpoint (default=False)')
optionalNamedcv.add_argument('-sv_mod', '--save_full_model', type = bool, required=False, default=False,
help='If -sv=True, whether to save the whole model or just weights (default=False, i.e. to save just weights); if -sv=False, nothing will be saved')
optionalNamedcv.add_argument('-cp_type', '--cp_save_type', type = str, required=False, default='last', help='Checkpoint saving options: best, last, everyN (default=last)')
optionalNamedcv.add_argument('-cp_n', '--cp_everyn_N', type = int, required=False, default=10, help='If -cp_type=everyN, the N value')
optionalNamedcv.add_argument('-v', '--verbose', type = bool, required=False, default=False, help='Display debug messages (default=False)')
parser_cv.set_defaults(func=truenet_tumseg_commands.cross_validate)
args = parser.parse_args()
if args.command == 'train':
truenet_tumseg_commands.train(args)
elif args.command == 'evaluate':
truenet_tumseg_commands.evaluate(args)
elif args.command == 'fine_tune':
truenet_tumseg_commands.fine_tune(args)
elif args.command == 'cross_validate':
truenet_tumseg_commands.cross_validate(args)
else:
parser.parse_args(["--help"])
sys.exit(0)
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment