Cylinder raytracing (#197)

Author: mcw92

.gitignore

@@ -174,7 +174,6 @@ examples/paint_plots/data/
 examples/paint_plots/plots/
 examples/paint_plots/results/
 examples/paint_plots/scenarios/
-examples/field_optimizations
 examples/field_optimizations/data_for_stral/
 examples/field_optimizations/measured_data/
 examples/field_optimizations/plots/

artist/core/blocking.py

@@ -27,12 +27,10 @@ def exponential(
     LRScheduler
         An exponential learning rate scheduler.
     """
-    scheduler = torch.optim.lr_scheduler.ExponentialLR(
+    return torch.optim.lr_scheduler.ExponentialLR(
         optimizer, gamma=float(parameters[config_dictionary.gamma])
     )
 
-    return scheduler
-
 
 def cyclic(
     optimizer: torch.optim.Optimizer,
@@ -53,15 +51,13 @@ def cyclic(
     LRScheduler
         A cyclic learning rate scheduler.
     """
-    scheduler = torch.optim.lr_scheduler.CyclicLR(
+    return torch.optim.lr_scheduler.CyclicLR(
         optimizer,
-        base_lr=float(parameters[config_dictionary.min]),
-        max_lr=float(parameters[config_dictionary.max]),
-        step_size_up=parameters[config_dictionary.step_size_up],
+        base_lr=float(parameters[config_dictionary.lr_min]),
+        max_lr=float(parameters[config_dictionary.lr_max]),
+        step_size_up=int(parameters[config_dictionary.step_size_up]),
     )
 
-    return scheduler
-
 
 def reduce_on_plateau(
     optimizer: torch.optim.Optimizer,
@@ -82,17 +78,15 @@ def reduce_on_plateau(
     LRScheduler
         A learning rate scheduler that reduces on plateaus.
     """
-    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(
+    return torch.optim.lr_scheduler.ReduceLROnPlateau(
         optimizer,
         factor=float(parameters[config_dictionary.reduce_factor]),
-        patience=parameters[config_dictionary.patience],
+        patience=int(parameters[config_dictionary.patience]),
         threshold=float(parameters[config_dictionary.threshold]),
-        cooldown=parameters[config_dictionary.cooldown],
-        min_lr=float(parameters[config_dictionary.min]),
+        cooldown=int(parameters[config_dictionary.cooldown]),
+        min_lr=float(parameters[config_dictionary.lr_min]),
     )
 
-    return scheduler
-
 
 class EarlyStopping:
     """

artist/core/heliostat_ray_tracer.py

@@ -19,7 +19,7 @@ class Loss:
 
     def __init__(self, loss_function: torch.nn.Module) -> None:
         """
-        Initialize the the base loss.
+        Initialize the base loss.
 
         Parameters
         ----------
@@ -41,10 +41,10 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of variable shape.
+            Shape is variable.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of variable shape.
+            Shape is variable.
         \*\*kwargs : Any
             Keyword arguments.
 
@@ -87,13 +87,13 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of shape [number_of_samples, ...].
+            Shape is ``[number_of_samples, ...]``.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of shape [number_of_samples, ...].
+            Shape is ``[number_of_samples, ...]``.
         \*\*kwargs : Any
             Keyword arguments.
-            The ``reduction_dimensions`` is an expected keyword argument for the vector loss.
+            ``reduction_dimensions`` is an expected keyword argument for the vector loss.
 
         Raises
         ------
@@ -104,7 +104,7 @@ def __call__(
         -------
         torch.Tensor
             The summed MSE vector loss reduced along the specified dimensions.
-            Tensor of shape [number_of_samples].
+            Shape is ``[number_of_samples]``.
         """
         expected_kwargs = ["reduction_dimensions"]
         for key in expected_kwargs:
@@ -162,13 +162,13 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of shape [number_of_samples, bitmap_resolution_e, bitmap_resolution_u].
+            Shape is ``[number_of_samples, bitmap_resolution_e, bitmap_resolution_u]``.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of shape [number_of_samples, 4].
+            Shape is ``[number_of_samples, 4]``.
         \*\*kwargs : Any
             Keyword arguments.
-            The ``reduction_dimensions``, ``target_area_indices`` and ``device`` are expected keyword arguments for the focal spot loss.
+            ``target_area_indices`` and ``device`` are expected keyword arguments for the focal spot loss.
 
         Raises
         ------
@@ -179,9 +179,9 @@ def __call__(
         -------
         torch.Tensor
             The focal spot loss.
-            Tensor of shape [number_of_samples].
+            Shape is ``[number_of_samples]``.
         """
-        expected_kwargs = ["reduction_dimensions", "device", "target_area_indices"]
+        expected_kwargs = ["device", "target_area_indices"]
         errors = []
         for key in expected_kwargs:
             if key not in kwargs:
@@ -196,21 +196,20 @@ def __call__(
 
         target_area_indices = kwargs["target_area_indices"]
 
-        focal_spot = utils.get_center_of_mass(
+        focal_spots_bitmap = utils.get_center_of_mass(
             bitmaps=prediction,
-            target_centers=self.scenario.target_areas.centers[target_area_indices],
-            target_widths=self.scenario.target_areas.dimensions[target_area_indices][
-                :, index_mapping.target_area_width
-            ],
-            target_heights=self.scenario.target_areas.dimensions[target_area_indices][
-                :, index_mapping.target_area_height
-            ],
             device=device,
         )
 
-        loss = torch.norm(focal_spot[:, :3] - ground_truth[:, :3], dim=1)
+        focal_spot_coordinates = utils.bitmap_coordinates_to_target_coordinates(
+            bitmap_coordinates=focal_spots_bitmap,
+            bitmap_resolution=torch.tensor(prediction.shape[1:]),
+            solar_tower=self.scenario.solar_tower,
+            target_area_indices=target_area_indices,
+            device=device,
+        )
 
-        return loss
+        return torch.norm(focal_spot_coordinates[:, :3] - ground_truth[:, :3], dim=1)
 
 
 class PixelLoss(Loss):
@@ -257,13 +256,13 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of shape [number_of_samples, bitmap_resolution_e, bitmap_resolution_u].
+            Shape is ``[number_of_samples, bitmap_resolution_e, bitmap_resolution_u]``.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of shape [number_of_samples, bitmap_resolution_e, bitmap_resolution_u].
+            Shape is ``[number_of_samples, bitmap_resolution_e, bitmap_resolution_u]``.
         \*\*kwargs : Any
             Keyword arguments.
-            The ``reduction_dimensions``, ``target_area_indices`` and optionally ``device`` are expected keyword arguments for the pixel loss.
+            ``reduction_dimensions``, ``target_area_indices``, and ``device`` are expected keyword arguments for the pixel loss.
 
         Raises
         ------
@@ -334,13 +333,13 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of shape [number_of_samples, bitmap_resolution_e, bitmap_resolution_u].
+            Shape is ``[number_of_samples, bitmap_resolution_e, bitmap_resolution_u]``.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of shape [number_of_samples, bitmap_resolution_e, bitmap_resolution_u].
+            Shape is ``[number_of_samples, bitmap_resolution_e, bitmap_resolution_u]``.
         \*\*kwargs : Any
             Keyword arguments.
-            The ``reduction_dimensions`` is an expected keyword argument for the KL-divergence loss.
+            ``reduction_dimensions`` is an expected keyword argument for the KL-divergence loss.
 
         Raises
         ------
@@ -351,7 +350,7 @@ def __call__(
         -------
         torch.Tensor
             The summed KL-divergence loss reduced along the specified dimensions.
-            Tensor of shape [number_of_samples].
+            Shape is ``[number_of_samples]``.
         """
         expected_kwargs = ["reduction_dimensions"]
         for key in expected_kwargs:
@@ -413,21 +412,17 @@ def __call__(
         ----------
         prediction : torch.Tensor
             The predicted values.
-            Tensor of shape [number_of_samples, 4].
+            Shape is ``[number_of_samples, 4]``.
         ground_truth : torch.Tensor
             The ground truth.
-            Tensor of shape [number_of_samples, 4].
+            Shape is ``[number_of_samples, 4]``.
         \*\*kwargs : Any
             Keyword arguments.
 
         Returns
         -------
         torch.Tensor
             The summed loss reduced along the specified dimensions.
-            Tensor of shape [number_of_samples].
+            Shape is ``[number_of_samples]``.
         """
-        cosine_similarity = self.loss_function(prediction, ground_truth)
-
-        loss = 1.0 - cosine_similarity
-
-        return loss
+        return 1.0 - self.loss_function(prediction, ground_truth)