applying the resize on a per-image basis, now. #48

mlcommons · Apr 27, 2021 · 91283ec · 91283ec
1 parent 83a54d8
commit 91283ec
Showing 1 changed file with 25 additions and 24 deletions.
diff --git a/GANDLF/data/ImagesFromDataFrame.py b/GANDLF/data/ImagesFromDataFrame.py
@@ -11,11 +11,9 @@
 from torchio import Image, Subject
 import SimpleITK as sitk
 # from GANDLF.utils import resize_image
-from GANDLF.preprocessing import (NonZeroNormalizeOnMaskedRegion, CropExternalZeroplanes,
-                                  resize_image_resolution, threshold_intensities,
+from GANDLF.preprocessing import (NonZeroNormalizeOnMaskedRegion, CropExternalZeroplanes, apply_resize, threshold_intensities,
                                   tensor_rotate_180, tensor_rotate_90, clip_intensities,
-                                  normalize_imagenet, normalize_standardize,
-                                  normalize_div_by_255)
+                                  normalize_imagenet, normalize_standardize, normalize_div_by_255)
 
 import copy, sys
 
@@ -163,10 +161,14 @@ def ImagesFromDataFrame(dataframe,
     for i in range(len(augmentation_patchAxesPoints)):
         augmentation_patchAxesPoints[i] = max(round(augmentation_patchAxesPoints[i] / 10), 1) # always at least have 1
 
-    # initialize resizeCheck if not present
-    if not('resizeCheck' in preprocessing):
-        preprocessing['resizeCheck'] = False
-
+    resize_images = False
+    # if resize has been defined but resample is not (or is none)
+    if not(preprocessing is None) and ('resize' in preprocessing):
+        if (preprocessing['resize'] is not None):
+            if not('resample' in preprocessing):
+                resize_images = True
+            else:
+                print('WARNING: \'resize\' is ignored as \'resample\' is defined under \'data_processing\', this will be skipped', file = sys.stderr)
     # iterating through the dataframe
     for patient in range(num_row):
         # We need this dict for storing the meta data for each subject
@@ -181,24 +183,17 @@ def ImagesFromDataFrame(dataframe,
             else:
                 img = sitk.ReadImage(str(dataframe[channel][patient]))
                 array = np.expand_dims(sitk.GetArrayFromImage(img), axis=0)
-                print("Image shape : ", img.shape, flush=True)
-                print("Array shape : ", array.shape, flush=True)
+                # print("Image shape : ", img.shape, flush=True)
+                # print("Array shape : ", array.shape, flush=True)
+                subject_dict[str(channel)] = Image(tensor=array, type=torchio.INTENSITY, path=dataframe[channel][patient])
+
+            # if resize is requested, the perform per-image resize with appropriate interpolator
+            if resize_images:
+                img = subject_dict[str(channel)].as_sitk()
+                img_resized = apply_resize(img, preprocessing_params=preprocessing)
+                array = np.expand_dims(sitk.GetArrayFromImage(img_resized), axis=0)
                 subject_dict[str(channel)] = Image(tensor=array, type=torchio.INTENSITY, path=dataframe[channel][patient])
 
-            # if resize has been defined but resample is not (or is none)
-            if not preprocessing['resizeCheck']:
-                if not(preprocessing is None) and ('resize' in preprocessing):
-                    if (preprocessing['resize'] is not None):
-                        preprocessing['resizeCheck'] = True
-                        if not('resample' in preprocessing):
-                            preprocessing['resample'] = {}
-                            if not('resolution' in preprocessing['resample']):
-                                preprocessing['resample']['resolution'] = resize_image_resolution(subject_dict[str(channel)].as_sitk(), preprocessing['resize'])
-                        else:
-                            print('WARNING: \'resize\' is ignored as \'resample\' is defined under \'data_processing\', this will be skipped', file = sys.stderr)
-                else:
-                    preprocessing['resizeCheck'] = True
-
         # # for regression
         # if predictionHeaders:
         #     # get the mask
@@ -213,6 +208,12 @@ def ImagesFromDataFrame(dataframe,
                 array = np.expand_dims(sitk.GetArrayFromImage(img), axis=0)
                 subject_dict['label'] = Image(tensor=array, type=torchio.LABEL, path=dataframe[labelHeader][patient])
 
+            # if resize is requested, the perform per-image resize with appropriate interpolator
+            if resize_images:
+                img = sitk.ReadImage(str(dataframe[labelHeader][patient]))
+                img_resized = apply_resize(img, preprocessing_params=preprocessing, interpolator=sitk.sitkNearestNeighbor)
+                array = np.expand_dims(sitk.GetArrayFromImage(img_resized), axis=0)
+                subject_dict['label'] = Image(tensor=array, type=torchio.LABEL, path=dataframe[channel][patient])
 
             subject_dict['path_to_metadata'] = str(dataframe[labelHeader][patient])
         else: