Merge branch 'master' into for_beluga_scilpy2

dwi_ml/data/dataset/multi_subject_containers.py

@@ -281,7 +281,6 @@ def load(self, hdf_handle: h5py.File, subj_id=None):
                     hdf_handle, subj_id, ref_group_info)
 
                 # Add subject to the list
-                logger.debug("     Adding it to the list of subjects.")
                 subj_idx = self.subjs_data_list.add_subject(subj_data)
 
                 # Arrange streamlines
@@ -290,7 +289,6 @@ def load(self, hdf_handle: h5py.File, subj_id=None):
                 if subj_data.is_lazy:
                     subj_data.add_handle(hdf_handle)
 
-                logger.debug("     Counting streamlines")
                 for group in range(len(self.streamline_groups)):
                     subj_sft_data = subj_data.sft_data_list[group]
                     n_streamlines = len(subj_sft_data)
@@ -302,6 +300,7 @@ def load(self, hdf_handle: h5py.File, subj_id=None):
                 subj_data.hdf_handle = None
 
             # Arrange final data properties: Concatenate all subjects
+            logging.debug("All subjects added. Final verifications.")
             self.streamline_lengths_mm = \
                 [np.concatenate(lengths_mm[group], axis=0)
                  for group in range(len(self.streamline_groups))]

dwi_ml/data/dataset/single_subject_containers.py

@@ -110,7 +110,8 @@ def init_single_subject_from_hdf(
             subject_mri_data_list.append(subject_mri_group_data)
 
         for group in streamline_groups:
-            logger.debug("        Loading subject's streamlines")
+            logger.debug("        Loading streamlines group '{}'"
+                         .format(group))
             sft_data = SFTData.init_sft_data_from_hdf_info(
                 hdf_file[subject_id][group])
             subject_sft_data_list.append(sft_data)

dwi_ml/data/hdf5/hdf5_creation.py

@@ -59,8 +59,8 @@ def format_filelist(filenames, enforce_presence, folder=None) -> List[str]:
     return new_files
 
 
-def _load_and_verify_file(filename: str, subj_input_path, group_name: str,
-                          group_affine, group_res):
+def _load_and_verify_file(filename: str, group_name: str, group_affine,
+                          group_res):
     """
     Loads a 3D or 4D nifti file. If it is a 3D dataset, adds a dimension to
     make it 4D. Then checks that it is compatible with a given group based on
@@ -70,26 +70,22 @@ def _load_and_verify_file(filename: str, subj_input_path, group_name: str,
     ------
     filename: str
         File's name. Must be .nii or .nii.gz.
-    subj_input_path: Path
-        Path where to load the nifti file from.
     group_name: str
         Name of the group with which 'filename' file must be compatible.
     group_affine: np.array
         The loaded file's affine must be equal (or very close) to this affine.
     group_res: np.array
         The loaded file's resolution must be equal (or very close) to this res.
     """
-    data_file = subj_input_path.joinpath(filename)
-
-    if not data_file.is_file():
+    if not os.path.isfile(filename):
         logging.debug("      Skipping file {} because it was not "
                       "found in this subject's folder".format(filename))
         # Note: if args.enforce_files_presence was set to true, this
         # case is not possible, already checked in
         # create_hdf5_dataset.py.
         return None
 
-    data, affine, res, _ = load_file_to4d(data_file)
+    data, affine, res, _ = load_file_to4d(filename)
 
     if not np.allclose(affine, group_affine, atol=1e-5):
         # Note. When keeping default options on tolerance, we have run
@@ -483,6 +479,7 @@ def _process_one_volume_group(self, group: str, subj_id: str,
                     else:
                         std_mask = np.logical_or(sub_mask_data, std_mask)
 
+        # Get the files and add the subject_dir as prefix.
         file_list = self.groups_config[group]['files']
         file_list = format_filelist(file_list, self.enforce_files_presence,
                                     folder=subj_input_dir)
@@ -505,8 +502,8 @@ def _process_one_volume_group(self, group: str, subj_id: str,
             for file_name in file_list[1:]:
                 logging.info("       - Processing file {}"
                              .format(os.path.basename(file_name)))
-                data = _load_and_verify_file(file_name, subj_input_dir, group,
-                                             group_affine, group_res)
+                data = _load_and_verify_file(file_name, group, group_affine,
+                                             group_res)
 
                 if std_option == 'per_file':
                     logging.info('          - Standardizing')

dwi_ml/data/processing/space/neighborhood.py

@@ -19,15 +19,17 @@ def prepare_neighborhood_vectors(
     neighborhood_type: str
         Either the 'axes' option or the 'grid' option. See each method for a
         description.
-    neighborhood_radius: Optional[int]
-        The radius. For the axes option: a radius of 1 = 7 neighbhors, a
-        radius of 2 = 13 . For the grid option: a radius of 1 = 27 neighbors,
-        a radius of 2 = 125.
-    neighborhood_resolution: Optional[float]
-        Resolution between each layer of neighborhood, in voxel space as compared
-        to the MRI data. Ex: 0.5 one neighborhood every half voxel.
-        Hint: Neighborhood's space will depend on this value. To convert from
-        mm to voxel world, you may use
+    neighborhood_radius: int
+        Required if neighborhood type is not None.
+        - axes: a radius of 1 = 7 neighbhors, a radius of 2 = 13, on two
+         concentric spheres.
+        - grid option: a radius of 1 = 27 neighbors, a radius of 2 = 125, on
+         two concentric cubes.
+    neighborhood_resolution: float
+        Required if neighborhood type is not None.
+        - 'axes': spacing between each concentric sphere.
+        - 'grid': resolution of the final grid-like neighborhood.
+        Hint: To convert from mm to voxel world, you may use
         dwi_ml.data.processing.space.world_to_vox.convert_world_to_vox(
             radius_mm, affine_mm_to_vox)
 
@@ -68,9 +70,11 @@ def prepare_neighborhood_vectors(
 
 def get_neighborhood_vectors_axes(radius: int, resolution: float):
     """
-    This neighborhood definition lies on a sphere. Returns a list of 7
-    positions (current, up, down, left, right, behind, in front) at exactly
-    `resolution` (mm or voxels) from origin (i.e. current postion).
+    This neighborhood definition lies on a sphere.
+
+    For radius = 1, returns a list of 7 positions (current, up, down, left,
+    right, behind, in front) at exactly `resolution` (mm or voxels) from origin
+    (i.e. current postion).
     If radius is > 1, returns a multi-radius neighborhood (lying on
     concentring spheres).
 

dwi_ml/experiment_utils/prints.py

@@ -2,15 +2,17 @@
 from collections.abc import Mapping
 
 
-def format_dict_to_str(d, indent=1):
+def format_dict_to_str(d, indent=1, keys_only=False):
     indentation = indent * "    "
     return ("\n" + indentation) + ("\n" + indentation).join(
-        "{!r}: {},".format(k, _format_val_to_str(v, indent+1))
+        "{!r}: {},".format(k, _format_val_to_str(v, indent+1, keys_only))
         for k, v in d.items())
 
 
-def _format_val_to_str(v, indent):
+def _format_val_to_str(v, indent, keys_only=False):
     if isinstance(v, Mapping):
-        return format_dict_to_str(v, indent)
+        return format_dict_to_str(v, indent, keys_only)
+    elif keys_only:
+        return ''
     else:
         return v