predict.py

"""
If you use this code, please cite the paper listed in:
https://github.com/BBillot/hypothalamus_seg/blob/master/bibtex.bib

Copyright 2020 Benjamin Billot

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in
compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is
distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied. See the License for the specific language governing permissions and limitations under the
License.
"""


from argparse import ArgumentParser
from hypothalamus_seg.predict import predict

parser = ArgumentParser()

# positional arguments
parser.add_argument("path_images", type=str,
                    help="images to segment. Can be the path to a single image or to a folder.")
parser.add_argument("path_segmentations", type=str,
                    help="saving path. Can be a single image (if path_images is an image), or a folder.")
parser.add_argument("path_model", type=str, help="path to a h5 file containing the trained model.")
parser.add_argument("segmentation_labels", type=str,
                    help="path to a numpy array containing all the segmentation label values")

# saving paths
parser.add_argument("--post", type=str, default=None, dest="path_posteriors",
                    help="path to a folder where to save the output posteriors")
parser.add_argument("--resampled", type=str, default=None, dest="path_resampled",
                    help="path to a csv file where to save the volumes of all subunits for all patients.")
parser.add_argument("--vol", type=str, default=None, dest="path_volumes",
                    help="path to a csv file where to save the volumes of all subunits for all patients.")

# preprocessing/postprocessing parameters
parser.add_argument("--padding", type=int, dest="padding", default=None,
                    help="margin to pad images with before prediction. Default is no padding.")
parser.add_argument("--cropping", type=int, dest="cropping", default=None,
                    help="crop images to this size before prediction. Default is no cropping.")
parser.add_argument("--target_res", type=float, dest="target_res", default=1.,
                    help="resolution at which we obtain predictions. Must be the same as the resolution of the "
                         "training data. Default is 1mm isotropic.")
parser.add_argument("--smoothing", type=float, dest="sigma_smoothing", default=0,
                    help="std dev of Gaussian kernel to smooth the predictions.")
parser.add_argument("--biggest_component", action='store_true', dest="keep_biggest_component",
                    help="whether to only keep the biggest component (linked voxels) of the segmentation")

# Architecture parameters (should be the same as training)
parser.add_argument("--conv_size", type=int, dest="conv_size", default=3, help="size of convolution masks")
parser.add_argument("--n_levels", type=int, dest="n_levels", default=3, help="number of level")
parser.add_argument("--conv_per_level", type=int, dest="nb_conv_per_level", default=2, help="convolution per level")
parser.add_argument("--unet_features", type=int, dest="unet_feat_count", default=24, help="features of first layer")
parser.add_argument("--feat_mult", type=int, dest="feat_multiplier", default=2, help="feature multiplication ratio")
parser.add_argument("--activation", type=str, dest="activation", default='elu', help="activation function")

# Evaluation parameters
parser.add_argument("--gt_folder", type=str, default=None, dest="gt_folder",
                    help="folder with ground truth segmentations for evaluation.")
parser.add_argument("--eval_label_list", type=str, dest="evaluation_labels", default=None,
                    help="labels to evaluate Dice scores on if gt is provided. Default is the same as label_list.")
parser.add_argument("--distances", action='store_true', dest="compute_distances",
                    help="whether to compute surface distances with gt segmentations. Default is False.")
parser.add_argument("--disable_whole_score", action='store_true', dest="compute_score_whole_structure",
                    help="whether to compute Dice scores for all foreground labels regrouped together.")

args = parser.parse_args()
predict(**vars(args))