#9 add losses and metrics

src/pixr/learning/loss.py

@@ -0,0 +1,25 @@
+import torch
+from typing import Optional
+from pixr.properties import LOSS_MSE
+
+
+def compute_loss(
+    predic: torch.Tensor,
+    target: torch.Tensor,
+    mode: Optional[str] = LOSS_MSE
+) -> torch.Tensor:
+    """
+    Compute loss based on the predicted and true values.
+
+    Args:
+        predic (torch.Tensor): [N, C, H, W] predicted values
+        target (torch.Tensor): [N, C, H, W] target values.
+        mode (Optional[str], optional): mode of loss computation.
+
+    Returns:
+        torch.Tensor: The computed loss.
+    """
+    assert mode in [LOSS_MSE], f"Mode {mode} not supported"
+    if mode == LOSS_MSE:
+        loss = torch.nn.functional.mse_loss(predic, target)
+    return loss

src/pixr/learning/metrics.py

@@ -0,0 +1,46 @@
+import torch
+from pixr.properties import METRIC_PSNR, REDUCTION_AVERAGE, REDUCTION_SKIP
+
+
+def compute_psnr(
+    predic: torch.Tensor,
+    target: torch.Tensor,
+    clamp_mse=1e-10,
+    reduction=REDUCTION_AVERAGE
+) -> torch.Tensor:
+    """
+    Compute the average PSNR metric for a batch of predicted and true values.
+
+    Args:
+        predic (torch.Tensor): [N, C, H, W] predicted values.
+        target (torch.Tensor): [N, C, H, W] target values.
+
+    Returns:
+        torch.Tensor: The average PSNR value for the batch.
+    """
+    mse_per_image = torch.mean((predic - target) ** 2, dim=(-3, -2, -1))
+    mse_per_image = torch.clamp(mse_per_image, min=clamp_mse)
+    psnr_per_image = 10 * torch.log10(1 / mse_per_image)
+    if reduction == REDUCTION_AVERAGE:
+        average_psnr = torch.mean(psnr_per_image)
+    elif reduction == REDUCTION_SKIP:
+        average_psnr = psnr_per_image
+    else:
+        raise ValueError(f"Unknown reduction {reduction}")
+    return average_psnr
+
+
+def compute_metrics(predic: torch.Tensor, target: torch.Tensor, reduction=REDUCTION_AVERAGE) -> dict:
+    """
+    Compute the metrics for a batch of predicted and true values.
+
+    Args:
+        predic (torch.Tensor): [N, C, H, W] predicted values.
+        target (torch.Tensor): [N, C, H, W] target values.
+
+    Returns:
+        dict: computed metrics.
+    """
+    average_psnr = compute_psnr(predic, target, reduction=reduction)
+    metrics = {METRIC_PSNR: average_psnr.item() if reduction != REDUCTION_SKIP else average_psnr}
+    return metrics

src/pixr/properties.py

@@ -8,3 +8,27 @@
 X = 0
 Y = 1
 Z = 2
+
+OUTPUT_FOLDER_NAME = "__output"
+DATALOADER = "data_loader"
+BATCH_SIZE = "batch_size"
+TRAIN, VALIDATION, TEST = "train", "validation", "test"
+ID = "id"
+NAME = "name"
+NB_EPOCHS = "nb_epochs"
+ARCHITECTURE = "architecture"
+MODEL = "model"
+NAME = "name"
+N_PARAMS = "n_params"
+OPTIMIZER = "optimizer"
+LR = "lr"
+PARAMS = "parameters"
+SCHEDULER_CONFIGURATION = "scheduler_configuration"
+SCHEDULER = "scheduler"
+REDUCELRONPLATEAU = "ReduceLROnPlateau"
+LOSS = "loss"
+LOSS_MSE = "mse"
+METRICS = "metrics"
+METRIC_PSNR = "psnr"
+REDUCTION_AVERAGE = "average"
+REDUCTION_SKIP = "skip"