Fix SnowMaskTask when dealing with empty patches

eolearn/coregistration/coregistration.py

@@ -114,7 +114,7 @@ def register(
                 warp_matrix,
                 warp_mode,
                 criteria,
-                valid_mask,  # type: ignore[arg-type]
+                valid_mask,
                 self.gauss_kernel_size,
             )
         except cv2.error as cv2err:  # pylint: disable=catching-non-exception

eolearn/features/utils.py

@@ -162,6 +162,7 @@ def spatially_resize_image(
     if resize_library is ResizeLib.CV2:
         resize_function = partial(cv2.resize, dsize=size, interpolation=resize_method.get_cv2_method(data.dtype))
     else:
+        # type: ignore[arg-type]
         resize_function = partial(_pil_resize_ndarray, size=size, method=resize_method.get_pil_method())
 
     resized_data = _apply_to_spatial_axes(resize_function, data, spatial_axes)

eolearn/mask/snow_mask.py

@@ -61,15 +61,12 @@ def _apply_dilation(self, snow_masks: np.ndarray) -> np.ndarray:
         """Apply binary dilation for each mask in the series"""
         if self.disk_size > 0:
             disk = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (self.disk_size, self.disk_size))
-            snow_masks = np.array([cv2.dilate(mask.astype(np.uint8), disk) for mask in snow_masks])
+            dilated_masks = np.array([cv2.dilate(mask.astype(np.uint8), disk) for mask in snow_masks])
+            snow_masks = dilated_masks.reshape(snow_masks.shape)  # edge case where data is empty
         return snow_masks.astype(bool)
 
     def execute(self, eopatch: EOPatch) -> EOPatch:
         bands = eopatch[self.bands_feature][..., self.band_indices]
-        if bands.shape[0] == 0:
-            eopatch[self.mask_feature] = np.zeros((*bands.shape[:-1], 1), dtype=bool)
-            return eopatch
-
         with np.errstate(divide="ignore", invalid="ignore"):
             # (B03 - B11) / (B03 + B11)
             ndsi = (bands[..., 0] - bands[..., 3]) / (bands[..., 0] + bands[..., 3])

tests/mask/test_snow_mask.py

@@ -10,7 +10,7 @@
 import numpy as np
 import pytest
 
-from eolearn.core import FeatureType
+from eolearn.core import FeatureType, EOPatch
 from eolearn.mask import SnowMaskTask
 
 
@@ -33,3 +33,12 @@ def test_snow_coverage(task, result, test_eopatch):
     snow_pixels = np.sum(output, axis=(1, 2, 3))
     assert np.sum(snow_pixels) == result[0], "Sum of snowy pixels does not match"
     assert snow_pixels[-4] == result[1], "Snowy pixels on specified frame do not match"
+
+
+def test_snow_empty_eopatch(test_eopatch):
+    bands = test_eopatch.data["BANDS-S2-L1C"]
+    empty_bands_array = np.array([], dtype=bands.dtype).reshape((0, *bands.shape[1:]))
+    empty_eopatch = EOPatch(bbox=test_eopatch.bbox, timestamps=[], data={"BANDS-S2-L1C": empty_bands_array})
+    task = SnowMaskTask((FeatureType.DATA, "BANDS-S2-L1C"), [2, 3, 7, 11], mask_name="TEST_SNOW_MASK")
+    resulting_eopatch = task(empty_eopatch) # checks if the task runs without errors
+    assert resulting_eopatch.mask["TEST_SNOW_MASK"].shape[0] == 0