Fix anomaly map shape to also work with tiling (#1959)

* Fix error message in Tiler

Signed-off-by: blaz-r <[email protected]>

* Fix anomaly map shapes to work with tiling

Signed-off-by: blaz-r <[email protected]>

---------

Signed-off-by: blaz-r <[email protected]>

Author: blaz-r

src/anomalib/data/utils/tiler.py

@@ -378,7 +378,7 @@ def tile(self, image: torch.Tensor, use_random_tiling: bool = False) -> torch.Te
         if self.input_h < self.tile_size_h or self.input_w < self.tile_size_w:
             msg = (
                 f"One of the edges of the tile size {self.tile_size_h, self.tile_size_w} is larger than "
-                f"that of the image {{self.input_h, self.input_w}}."
+                f"that of the image {self.input_h, self.input_w}."
             )
             raise ValueError(
                 msg,

src/anomalib/models/image/padim/torch_model.py

@@ -126,6 +126,7 @@ def forward(self, input_tensor: torch.Tensor) -> torch.Tensor:
             torch.Size([32, 128, 28, 28]),
             torch.Size([32, 256, 14, 14])]
         """
+        output_size = input_tensor.shape[-2:]
         if self.tiler:
             input_tensor = self.tiler.tile(input_tensor)
 
@@ -143,7 +144,7 @@ def forward(self, input_tensor: torch.Tensor) -> torch.Tensor:
                 embedding=embeddings,
                 mean=self.gaussian.mean,
                 inv_covariance=self.gaussian.inv_covariance,
-                image_size=input_tensor.shape[-2:],
+                image_size=output_size,
             )
         return output
 

src/anomalib/models/image/patchcore/torch_model.py

@@ -70,6 +70,7 @@ def forward(self, input_tensor: torch.Tensor) -> torch.Tensor | dict[str, torch.
         Returns:
             Tensor | dict[str, torch.Tensor]: Embedding for training, anomaly map and anomaly score for testing.
         """
+        output_size = input_tensor.shape[-2:]
         if self.tiler:
             input_tensor = self.tiler.tile(input_tensor)
 
@@ -98,7 +99,7 @@ def forward(self, input_tensor: torch.Tensor) -> torch.Tensor | dict[str, torch.
             # reshape to w, h
             patch_scores = patch_scores.reshape((batch_size, 1, width, height))
             # get anomaly map
-            anomaly_map = self.anomaly_map_generator(patch_scores, input_tensor.shape[-2:])
+            anomaly_map = self.anomaly_map_generator(patch_scores, output_size)
 
             output = {"anomaly_map": anomaly_map, "pred_score": pred_score}
 

src/anomalib/models/image/stfpm/torch_model.py

@@ -61,6 +61,7 @@ def forward(self, images: torch.Tensor) -> torch.Tensor | dict[str, torch.Tensor
         Returns:
           Teacher and student features when in training mode, otherwise the predicted anomaly maps.
         """
+        output_size = images.shape[-2:]
         if self.tiler:
             images = self.tiler.tile(images)
         teacher_features: dict[str, torch.Tensor] = self.teacher_model(images)
@@ -78,7 +79,7 @@ def forward(self, images: torch.Tensor) -> torch.Tensor | dict[str, torch.Tensor
             output = self.anomaly_map_generator(
                 teacher_features=teacher_features,
                 student_features=student_features,
-                image_size=images.shape[-2:],
+                image_size=output_size,
             )
 
         return output