Fix resized image masking

emerald/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '0.2.0'
+__version__ = '0.2.1'
 
 DISPLAY_TITLE = r"""
        _                                     _     _ 

emerald/emerald.py

@@ -101,15 +101,16 @@ def emerald(model: Unet, input_path: str, mask_path: Optional[Path], brain_paths
             post_processing: bool, footprint: Optional[npt.NDArray]):
     img_path = str(input_path)
 
-    img, hdr = getImageData(img_path)
+    img_original, hdr = getImageData(img_path)
+    img_resized = img_original
     resizeNeeded = False
 
-    if img.shape[1] != 256 or img.shape[2] != 256:
-        original_shape = (img.shape[2], img.shape[1])
-        img = __resizeData(img)
+    if img_original.shape[1] != 256 or img_original.shape[2] != 256:
+        original_shape = (img_original.shape[2], img_original.shape[1])
+        img_resized = __resizeData(img_original)
         resizeNeeded = True
 
-    res = model.predict_mask(img)
+    res = model.predict_mask(img_resized)
 
     if post_processing:
         res = __postProcessing(res, no_dilation=(footprint is not None), footprint=footprint)
@@ -130,12 +131,11 @@ def emerald(model: Unet, input_path: str, mask_path: Optional[Path], brain_paths
 
     if brain_paths:
         # for whatever reason, img.shape=(38, 256, 256, 1).
-        if len(img.shape) == 4 and img.shape[3] == 1:
-            img = np.squeeze(img)
-            img = np.moveaxis(img, 0, -1)
+        if len(img_original.shape) == 4 and img_original.shape[3] == 1:
+            img_original = np.squeeze(img_original)
+            img_original = np.moveaxis(img_original, 0, -1)
 
     # apply res mask to img
     for mult, brain_path in brain_paths:
-        print(f'img.shape={img.shape}, res.shape={res.shape}')
-        overlayed_data = np.clip(res, mult, 1.0) * img
+        overlayed_data = np.clip(res, mult, 1.0) * img_original
         save(overlayed_data, str(brain_path), hdr)