refactor: Renamed patch_offset to step_size

README.md

@@ -29,7 +29,7 @@ You can install `patchly` via [pip](https://pypi.org/project/patchly/):
 
 Demonstration on how to use Patchly for sliding-window patchification and subsequent aggregation:
 ```python
-sampler = GridSampler(spatial_size, patch_size, patch_offset, image)
+sampler = GridSampler(spatial_size, patch_size, step_size, image)
 aggregator = Aggregator(sampler, output_size)
 
 for patch, patch_bbox in sampler:
@@ -67,7 +67,7 @@ def model(x):
     return y
 
 # Init GridSampler
-sampler = GridSampler(image=np.random.random((1000, 1000, 3)), spatial_size=(1000, 1000), patch_size=(100, 100), patch_offset=(50, 50))
+sampler = GridSampler(image=np.random.random((1000, 1000, 3)), spatial_size=(1000, 1000), patch_size=(100, 100), step_size=(50, 50))
 # Init dataloader
 loader = DataLoader(ExampleDataset(sampler), batch_size=4, num_workers=0, shuffle=False)
 # Init aggregator

examples/inference_chunk.py

@@ -12,11 +12,11 @@ def example():
     image = np.random.random((3, 1000, 1000))  # Channel, Width, Height
     spatial_size = image.shape[-2:]
     patch_size = (100, 100)
-    patch_offset = (50, 50)
+    step_size = (50, 50)
     chunk_size = (500, 500)
 
     # Init GridSampler
-    sampler = GridSampler(image=image, spatial_size=spatial_size, patch_size=patch_size, patch_offset=patch_offset, spatial_first=False)
+    sampler = GridSampler(image=image, spatial_size=spatial_size, patch_size=patch_size, step_size=step_size, spatial_first=False)
     # Convert sampler into a PyTorch dataset
     loader = SamplerDataset(sampler)
     # Init dataloader

examples/pytorch_inference.py

@@ -7,7 +7,7 @@
 
 def example():
     # Init GridSampler
-    sampler = GridSampler(image=np.random.random((1000, 1000, 3)), spatial_size=(1000, 1000), patch_size=(100, 100), patch_offset=(50, 50))
+    sampler = GridSampler(image=np.random.random((1000, 1000, 3)), spatial_size=(1000, 1000), patch_size=(100, 100), step_size=(50, 50))
     # Init dataloader
     loader = DataLoader(ExampleDataset(sampler), batch_size=4, num_workers=0, shuffle=False)
     # Init aggregator

misc/tmp.py

@@ -5,9 +5,9 @@
 import copy
 
 
-def _test_aggregator(image, spatial_size, patch_size, patch_offset=None, spatial_first_sampler=True, spatial_first_aggregator=True, output=None, weights='avg', multiply_patch_by_index=False, softmax_dim=None):        
+def _test_aggregator(image, spatial_size, patch_size, step_size=None, spatial_first_sampler=True, spatial_first_aggregator=True, output=None, weights='avg', multiply_patch_by_index=False, softmax_dim=None):        
         # Test with output size
-        sampler = GridSampler(image=copy.deepcopy(image), spatial_size=spatial_size, patch_size=patch_size, patch_offset=patch_offset, spatial_first=spatial_first_sampler, mode=SamplingMode.SAMPLE_SQUEEZE)
+        sampler = GridSampler(image=copy.deepcopy(image), spatial_size=spatial_size, patch_size=patch_size, step_size=step_size, spatial_first=spatial_first_sampler, mode=SamplingMode.SAMPLE_SQUEEZE)
         aggregator = Aggregator(sampler=sampler, output_size=image.shape, weights=weights, spatial_first=spatial_first_aggregator, softmax_dim=softmax_dim)
 
         for i, (patch, patch_bbox) in enumerate(sampler):
@@ -28,7 +28,7 @@ def _test_aggregator(image, spatial_size, patch_size, patch_offset=None, spatial
         # Test without output array
         if output is None:
             output = np.zeros_like(image)
-        sampler = GridSampler(image=copy.deepcopy(image), spatial_size=spatial_size, patch_size=patch_size, patch_offset=patch_offset, spatial_first=spatial_first_sampler, mode=SamplingMode.SAMPLE_SQUEEZE)
+        sampler = GridSampler(image=copy.deepcopy(image), spatial_size=spatial_size, patch_size=patch_size, step_size=step_size, spatial_first=spatial_first_sampler, mode=SamplingMode.SAMPLE_SQUEEZE)
         aggregator = Aggregator(sampler=sampler, output=output, weights=weights, spatial_first=spatial_first_aggregator, softmax_dim=softmax_dim)
 
         for i, (patch, patch_bbox) in enumerate(sampler):
@@ -48,9 +48,9 @@ def _test_aggregator(image, spatial_size, patch_size, patch_offset=None, spatial
 
         return output1, output2
 
-patch_offset = (3, 3)
+step_size = (3, 3)
 patch_size = (10, 10)
 spatial_size = (103, 107)
 image = np.random.random(spatial_size)
 
-_test_aggregator(image, spatial_size, patch_size, patch_offset=patch_offset)
+_test_aggregator(image, spatial_size, patch_size, step_size=step_size)

patchly/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.0.4"
+__version__ = "0.0.5"
 
 from patchly.sampler import GridSampler, SamplingMode
 from patchly.aggregator import Aggregator

patchly/aggregator.py

@@ -29,7 +29,7 @@ def __init__(self, sampler: GridSampler, output_size: Optional[Union[Tuple, npt.
         self.sampler = sampler
         self.image_size_s = sampler.image_size_s
         self.patch_size_s = sampler.patch_size_s
-        self.patch_offset_s = sampler.patch_offset_s
+        self.step_size_s = sampler.step_size_s
         self.chunk_size_s = chunk_size
         self.spatial_first = spatial_first
         self.mode = sampler.mode
@@ -114,7 +114,7 @@ def set_aggregator(self, sampler, output_h, softmax_dim):
                                   output_h=output_h, spatial_first=self.spatial_first, softmax_dim=softmax_dim, has_batch_dim=self.has_batch_dim, 
                                   weight_patch_s=self.weight_patch_s, weight_map_s=self.weight_map_s, device=self.device, array_type=self.array_type)
         elif self.mode.name.startswith('SAMPLE_') and self.chunk_size_s is not None:
-            aggregator = _ChunkAggregator(sampler=sampler, image_size_s=self.image_size_s, patch_size_s=self.patch_size_s, patch_offset_s=self.patch_offset_s,
+            aggregator = _ChunkAggregator(sampler=sampler, image_size_s=self.image_size_s, patch_size_s=self.patch_size_s, step_size_s=self.step_size_s,
                                        chunk_size_s=self.chunk_size_s, output_h=output_h, spatial_first=self.spatial_first, softmax_dim=softmax_dim, 
                                        has_batch_dim=self.has_batch_dim, weight_patch_s=self.weight_patch_s, device=self.device, array_type=self.array_type)
         elif self.mode.name.startswith('PAD_') and self.chunk_size_s is None:
@@ -263,7 +263,7 @@ def verify_array_types(self, array_type):
 
 
 class _ChunkAggregator(_Aggregator):
-    def __init__(self, sampler: GridSampler, image_size_s: Union[Tuple, npt.ArrayLike], patch_size_s: Union[Tuple, npt.ArrayLike], patch_offset_s: Union[Tuple, npt.ArrayLike], chunk_size_s: Union[Tuple, npt.ArrayLike],
+    def __init__(self, sampler: GridSampler, image_size_s: Union[Tuple, npt.ArrayLike], patch_size_s: Union[Tuple, npt.ArrayLike], step_size_s: Union[Tuple, npt.ArrayLike], chunk_size_s: Union[Tuple, npt.ArrayLike],
                  output_h: Optional[npt.ArrayLike] = None, spatial_first: bool = True,
                  softmax_dim: Optional[int] = None, has_batch_dim: bool = False, weight_patch_s: npt.ArrayLike = None, device = 'cpu', array_type = None):
         """
@@ -278,7 +278,7 @@ def __init__(self, sampler: GridSampler, image_size_s: Union[Tuple, npt.ArrayLik
         """
         super().__init__(sampler=sampler, image_size_s=image_size_s, patch_size_s=patch_size_s, output_h=output_h, spatial_first=spatial_first, 
                          softmax_dim=softmax_dim, has_batch_dim=has_batch_dim, weight_patch_s=weight_patch_s, weight_map_s=None, device=device, array_type=array_type)
-        self.patch_offset_s = patch_offset_s
+        self.step_size_s = step_size_s
         self.chunk_size_s = chunk_size_s
         self.chunk_dtype = self.set_chunk_dtype()
         self.chunk_sampler, self.chunk_patch_dict, self.patch_chunk_dict = self.compute_patches()
@@ -292,7 +292,7 @@ def set_chunk_dtype(self):
 
     def compute_patches(self):
         patch_sampler = self.sampler
-        chunk_sampler = _AdaptiveGridSampler(image_size_s=self.image_size_s, patch_size_s=self.chunk_size_s, patch_offset_s=self.chunk_size_s)
+        chunk_sampler = _AdaptiveGridSampler(image_size_s=self.image_size_s, patch_size_s=self.chunk_size_s, step_size_s=self.chunk_size_s)
         chunk_patch_dict = defaultdict(dict)
         patch_chunk_dict = defaultdict(dict)