Address tiler issues (#1411)

* fix tiler

* deprecate random tile locations

* restore random tiling in tile method

* check tiling section in config

* disable tiling for ganomalu

* pad -> padding

notebooks/100_datamodules/104_tiling.ipynb

@@ -214,7 +214,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "ToPILImage()(make_grid(tiled_image, nrow=4, pad=5))"
+    "ToPILImage()(make_grid(tiled_image, nrow=4, padding=5))"
    ]
   },
   {

src/anomalib/models/ganomaly/config.yaml

@@ -18,13 +18,6 @@ dataset:
   test_split_ratio: 0.2 # fraction of train images held out testing (usage depends on test_split_mode)
   val_split_mode: same_as_test # options: [same_as_test, from_test, synthetic]
   val_split_ratio: 0.5 # fraction of train/test images held out for validation (usage depends on val_split_mode)
-  tiling:
-    apply: true
-    tile_size: 64
-    stride: null
-    remove_border_count: 0
-    use_random_tiling: False
-    random_tile_count: 16
 
 model:
   name: ganomaly

src/anomalib/pre_processing/tiler.py

@@ -160,10 +160,9 @@ def __init__(
         stride: int | Sequence | None = None,
         remove_border_count: int = 0,
         mode: ImageUpscaleMode = ImageUpscaleMode.PADDING,
-        tile_count: int = 4,
     ) -> None:
         self.tile_size_h, self.tile_size_w = self.__validate_size_type(tile_size)
-        self.tile_count = tile_count
+        self.random_tile_count = 4
 
         if stride is not None:
             self.stride_h, self.stride_w = self.__validate_size_type(stride)
@@ -218,7 +217,7 @@ def __random_tile(self, image: Tensor) -> Tensor:
 
         Returns: Randomly cropped tiles from the image
         """
-        return torch.vstack([T.RandomCrop(self.tile_size_h)(image) for i in range(self.tile_count)])
+        return torch.vstack([T.RandomCrop(self.tile_size_h)(image) for i in range(self.random_tile_count)])
 
     def __unfold(self, tensor: Tensor) -> Tensor:
         """Unfolds tensor into tiles.
@@ -339,7 +338,7 @@ def __fold(self, tiles: Tensor) -> Tensor:
 
         return img
 
-    def tile(self, image: Tensor, use_random_tiling: bool | None = False) -> Tensor:
+    def tile(self, image: Tensor, use_random_tiling: bool = False) -> Tensor:
         """Tiles an input image to either overlapping, non-overlapping or random patches.
 
         Args:

src/anomalib/utils/callbacks/__init__.py

@@ -88,6 +88,17 @@ def get_callbacks(config: DictConfig | ListConfig) -> list[Callback]:
     )
     callbacks.append(post_processing_callback)
 
+    # Add the tiling configuration callback
+    if config.dataset.get("tiling") and config.dataset.tiling.apply:
+        callbacks.append(
+            TilerConfigurationCallback(
+                enable=True,
+                tile_size=config.dataset.tiling.tile_size,
+                stride=config.dataset.tiling.stride,
+                remove_border_count=config.dataset.tiling.remove_border_count,
+            )
+        )
+
     # Add metric configuration to the model via MetricsConfigurationCallback
     metrics_callback = MetricsConfigurationCallback(
         config.dataset.task,

src/anomalib/utils/callbacks/tiler_configuration.py

@@ -27,7 +27,6 @@ def __init__(
         stride: int | Sequence | None = None,
         remove_border_count: int = 0,
         mode: ImageUpscaleMode = ImageUpscaleMode.PADDING,
-        tile_count: int = 4,
     ) -> None:
         """Sets tiling configuration from the command line.
 
@@ -49,7 +48,6 @@ def __init__(
         self.stride = stride
         self.remove_border_count = remove_border_count
         self.mode = mode
-        self.tile_count = tile_count
 
     def setup(self, trainer: pl.Trainer, pl_module: pl.LightningModule, stage: str | None = None) -> None:
         """Setup Tiler object within Anomalib Model.
@@ -72,7 +70,6 @@ def setup(self, trainer: pl.Trainer, pl_module: pl.LightningModule, stage: str |
                     stride=self.stride,
                     remove_border_count=self.remove_border_count,
                     mode=self.mode,
-                    tile_count=self.tile_count,
                 )
             else:
                 raise ValueError("Model does not support tiling.")