Fix error in ergas calculation (#2498)

* fix error in formula

* fix doctests

* changelog

* fix other doctests

Author: SkafteNicki

CHANGELOG.md

@@ -36,6 +36,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Fixed axis names with Precision-Recall curve ([#2462](https://github.com/Lightning-AI/torchmetrics/pull/2462))
 
 
+- Fixed bug in computation of `ERGAS` metric ([#2498](https://github.com/Lightning-AI/torchmetrics/pull/2498))
+
+
 - Fixed `BootStrapper` wrapper not working with `kwargs` provided argument ([#2503](https://github.com/Lightning-AI/torchmetrics/pull/2503))
 
 

src/torchmetrics/functional/image/_deprecated.py

@@ -53,7 +53,7 @@ def _error_relative_global_dimensionless_synthesis(
     >>> target = preds * 0.75
     >>> ergds = _error_relative_global_dimensionless_synthesis(preds, target)
     >>> torch.round(ergds)
-    tensor(154.)
+    tensor(10.)
 
     """
     _deprecated_root_import_func("error_relative_global_dimensionless_synthesis", "image")

src/torchmetrics/functional/image/ergas.py

@@ -67,7 +67,7 @@ def _ergas_compute(
         >>> target = preds * 0.75
         >>> preds, target = _ergas_update(preds, target)
         >>> torch.round(_ergas_compute(preds, target))
-        tensor(154.)
+        tensor(10.)
 
     """
     b, c, h, w = preds.shape
@@ -79,7 +79,7 @@ def _ergas_compute(
     rmse_per_band = torch.sqrt(sum_squared_error / (h * w))
     mean_target = torch.mean(target, dim=2)
 
-    ergas_score = 100 * ratio * torch.sqrt(torch.sum((rmse_per_band / mean_target) ** 2, dim=1) / c)
+    ergas_score = 100 / ratio * torch.sqrt(torch.sum((rmse_per_band / mean_target) ** 2, dim=1) / c)
     return reduce(ergas_score, reduction)
 
 
@@ -115,9 +115,8 @@ def error_relative_global_dimensionless_synthesis(
         >>> gen = torch.manual_seed(42)
         >>> preds = torch.rand([16, 1, 16, 16], generator=gen)
         >>> target = preds * 0.75
-        >>> ergds = error_relative_global_dimensionless_synthesis(preds, target)
-        >>> torch.round(ergds)
-        tensor(154.)
+        >>> error_relative_global_dimensionless_synthesis(preds, target)
+        tensor(9.6193)
 
     """
     preds, target = _ergas_update(preds, target)

src/torchmetrics/image/_deprecated.py

@@ -22,7 +22,7 @@ class _ErrorRelativeGlobalDimensionlessSynthesis(ErrorRelativeGlobalDimensionles
     >>> target = preds * 0.75
     >>> ergas = _ErrorRelativeGlobalDimensionlessSynthesis()
     >>> torch.round(ergas(preds, target))
-    tensor(154.)
+    tensor(10.)
 
     """
 

src/torchmetrics/image/ergas.py

@@ -69,7 +69,7 @@ class ErrorRelativeGlobalDimensionlessSynthesis(Metric):
         >>> target = preds * 0.75
         >>> ergas = ErrorRelativeGlobalDimensionlessSynthesis()
         >>> torch.round(ergas(preds, target))
-        tensor(154.)
+        tensor(10.)
 
     """
 

tests/unittests/image/test_ergas.py

@@ -65,7 +65,7 @@ def _reference_ergas(
     rmse_per_band = torch.sqrt(sum_squared_error / (h * w))
     mean_target = torch.mean(sk_target, dim=2)
     # compute ergas score
-    ergas_score = 100 * ratio * torch.sqrt(torch.sum((rmse_per_band / mean_target) ** 2, dim=1) / c)
+    ergas_score = 100 / ratio * torch.sqrt(torch.sum((rmse_per_band / mean_target) ** 2, dim=1) / c)
     # reduction
     if reduction == "sum":
         return torch.sum(ergas_score)