Metrics

pl_bolts/metrics/object_detection.py

@@ -1,10 +1,7 @@
 import torch
 from torch import Tensor
 
-from pl_bolts.utils.stability import under_review
 
-
-@under_review()
 def iou(preds: Tensor, target: Tensor) -> Tensor:
     """Calculates the intersection over union.
 
@@ -33,11 +30,10 @@ def iou(preds: Tensor, target: Tensor) -> Tensor:
     pred_area = (preds[:, 2] - preds[:, 0]) * (preds[:, 3] - preds[:, 1])
     target_area = (target[:, 2] - target[:, 0]) * (target[:, 3] - target[:, 1])
     union = pred_area[:, None] + target_area - intersection
-    iou = torch.true_divide(intersection, union)
-    return iou
+    iou_value = torch.true_divide(intersection, union)
+    return iou_value
 
 
-@under_review()
 def giou(preds: Tensor, target: Tensor) -> Tensor:
     """Calculates the generalized intersection over union.
 
@@ -74,6 +70,6 @@ def giou(preds: Tensor, target: Tensor) -> Tensor:
     C_x_max = torch.max(preds[:, None, 2], target[:, 2])
     C_y_max = torch.max(preds[:, None, 3], target[:, 3])
     C_area = (C_x_max - C_x_min).clamp(min=0) * (C_y_max - C_y_min).clamp(min=0)
-    iou = torch.true_divide(intersection, union)
-    giou = iou - torch.true_divide((C_area - union), C_area)
-    return giou
+    iou_value = torch.true_divide(intersection, union)
+    giou_value = iou_value - torch.true_divide((C_area - union), C_area)
+    return giou_value

tests/metrics/test_object_detection.py

@@ -11,7 +11,7 @@
     [(torch.tensor([[100, 100, 200, 200]]), torch.tensor([[100, 100, 200, 200]]), torch.tensor([[1.0]]))],
 )
 def test_iou_complete_overlap(preds, target, expected_iou):
-    torch.testing.assert_allclose(iou(preds, target), expected_iou)
+    torch.testing.assert_close(iou(preds, target), expected_iou)
 
 
 @pytest.mark.parametrize(
@@ -22,7 +22,7 @@ def test_iou_complete_overlap(preds, target, expected_iou):
     ],
 )
 def test_iou_no_overlap(preds, target, expected_iou):
-    torch.testing.assert_allclose(iou(preds, target), expected_iou)
+    torch.testing.assert_close(iou(preds, target), expected_iou)
 
 
 @pytest.mark.parametrize(
@@ -36,15 +36,15 @@ def test_iou_no_overlap(preds, target, expected_iou):
     ],
 )
 def test_iou_multi(preds, target, expected_iou):
-    torch.testing.assert_allclose(iou(preds, target), expected_iou)
+    torch.testing.assert_close(iou(preds, target), expected_iou)
 
 
 @pytest.mark.parametrize(
     "preds, target, expected_giou",
     [(torch.tensor([[100, 100, 200, 200]]), torch.tensor([[100, 100, 200, 200]]), torch.tensor([[1.0]]))],
 )
 def test_complete_overlap(preds, target, expected_giou):
-    torch.testing.assert_allclose(giou(preds, target), expected_giou)
+    torch.testing.assert_close(giou(preds, target), expected_giou)
 
 
 @pytest.mark.parametrize(
@@ -55,7 +55,7 @@ def test_complete_overlap(preds, target, expected_giou):
     ],
 )
 def test_no_overlap(preds, target, expected_giou):
-    torch.testing.assert_allclose(giou(preds, target), expected_giou)
+    torch.testing.assert_close(giou(preds, target), expected_giou)
 
 
 @pytest.mark.parametrize(
@@ -69,4 +69,4 @@ def test_no_overlap(preds, target, expected_giou):
     ],
 )
 def test_giou_multi(preds, target, expected_giou):
-    torch.testing.assert_allclose(giou(preds, target), expected_giou)
+    torch.testing.assert_close(giou(preds, target), expected_giou, atol=0.0001, rtol=0.0001)