Store masks as bools

pathml/core/h5managers.py

@@ -157,7 +157,11 @@ def add_tile(self, tile):
  self.h5["tiles"][str(tile.coords)]["masks"].create_dataset(
  str(key),
  data=mask,
- dtype="float16",
+ chunks=True,
+ compression="gzip",
+ compression_opts=9, # masks should be highly compressible
+ shuffle=True,
+ dtype="bool",
  )
 
  # add coords

tests/conftest.py

@@ -30,7 +30,7 @@ def create_HE_tile():
  im_np_rgb = cv2.cvtColor(im_np, cv2.COLOR_RGBA2RGB)
  # make mask object
  masks = np.random.randint(
- low=1, high=255, size=(im_np_rgb.shape[0], im_np_rgb.shape[1]), dtype=np.uint8
+ low=0, high=1, size=(im_np_rgb.shape[0], im_np_rgb.shape[1]), dtype=bool
  )
  masks = {"testmask": masks}
  # labels dict
@@ -74,7 +74,7 @@ def tileVectra():
 
  # make mask object
  masks = np.random.randint(
- low=1, high=255, size=(region.shape[0], region.shape[1]), dtype=np.uint8
+ low=0, high=1, size=(region.shape[0], region.shape[1]), dtype=bool
  )
  masks = {"testmask": masks}
 

tests/core_tests/test_h5managers.py

@@ -49,12 +49,12 @@ def test_tile_dtype_HE(tileHE):
  slidedata.tiles.add(tileHE)
  tile_retrieved = slidedata.tiles[tileHE.coords]
  assert tile_retrieved.image.dtype == np.float16
- assert tile_retrieved.masks["testmask"].dtype == np.float16
+ assert tile_retrieved.masks["testmask"].dtype == bool
 
 
 def test_tile_dtype_IF(tileVectra, vectra_slide):
  """make sure that retrieved tiles and corresponding masks are float16"""
  vectra_slide.tiles.add(tileVectra)
  tile_retrieved = vectra_slide.tiles[tileVectra.coords]
  assert tile_retrieved.image.dtype == np.float16
- assert tile_retrieved.masks["testmask"].dtype == np.float16
+ assert tile_retrieved.masks["testmask"].dtype == bool