diff --git a/torchvision/datasets/vision.py b/torchvision/datasets/vision.py
index 22fc85322e4..dad19f65197 100644
--- a/torchvision/datasets/vision.py
+++ b/torchvision/datasets/vision.py
@@ -28,6 +28,7 @@ class VisionDataset(data.Dataset):
 
     _repr_indent = 4
 
+    @_log_api_usage_once
     def __init__(
         self,
         root: str,
@@ -35,7 +36,6 @@ def __init__(
         transform: Optional[Callable] = None,
         target_transform: Optional[Callable] = None,
     ) -> None:
-        _log_api_usage_once(self)
         if isinstance(root, torch._six.string_classes):
             root = os.path.expanduser(root)
         self.root = root
diff --git a/torchvision/models/alexnet.py b/torchvision/models/alexnet.py
index bb812febdc4..2a6e2ccd7fa 100644
--- a/torchvision/models/alexnet.py
+++ b/torchvision/models/alexnet.py
@@ -16,9 +16,9 @@
 
 
 class AlexNet(nn.Module):
+    @_log_api_usage_once
     def __init__(self, num_classes: int = 1000, dropout: float = 0.5) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.features = nn.Sequential(
             nn.Conv2d(3, 64, kernel_size=11, stride=4, padding=2),
             nn.ReLU(inplace=True),
diff --git a/torchvision/models/densenet.py b/torchvision/models/densenet.py
index 14e318360af..96e5b3d1fb8 100644
--- a/torchvision/models/densenet.py
+++ b/torchvision/models/densenet.py
@@ -151,6 +151,7 @@ class DenseNet(nn.Module):
           but slower. Default: *False*. See `"paper" <https://arxiv.org/pdf/1707.06990.pdf>`_.
     """
 
+    @_log_api_usage_once
     def __init__(
         self,
         growth_rate: int = 32,
@@ -161,9 +162,7 @@ def __init__(
         num_classes: int = 1000,
         memory_efficient: bool = False,
     ) -> None:
-
         super().__init__()
-        _log_api_usage_once(self)
 
         # First convolution
         self.features = nn.Sequential(
diff --git a/torchvision/models/detection/generalized_rcnn.py b/torchvision/models/detection/generalized_rcnn.py
index 37ef1820d71..53a0402f870 100644
--- a/torchvision/models/detection/generalized_rcnn.py
+++ b/torchvision/models/detection/generalized_rcnn.py
@@ -25,9 +25,9 @@ class GeneralizedRCNN(nn.Module):
             the model
     """
 
+    @_log_api_usage_once
     def __init__(self, backbone: nn.Module, rpn: nn.Module, roi_heads: nn.Module, transform: nn.Module) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.transform = transform
         self.backbone = backbone
         self.rpn = rpn
diff --git a/torchvision/models/detection/retinanet.py b/torchvision/models/detection/retinanet.py
index e5ced9870ba..fa0eb4f4da1 100644
--- a/torchvision/models/detection/retinanet.py
+++ b/torchvision/models/detection/retinanet.py
@@ -316,6 +316,7 @@ class RetinaNet(nn.Module):
         "proposal_matcher": det_utils.Matcher,
     }
 
+    @_log_api_usage_once
     def __init__(
         self,
         backbone,
@@ -337,7 +338,6 @@ def __init__(
         topk_candidates=1000,
     ):
         super().__init__()
-        _log_api_usage_once(self)
 
         if not hasattr(backbone, "out_channels"):
             raise ValueError(
diff --git a/torchvision/models/detection/ssd.py b/torchvision/models/detection/ssd.py
index 5778a07075d..d14119777e2 100644
--- a/torchvision/models/detection/ssd.py
+++ b/torchvision/models/detection/ssd.py
@@ -165,6 +165,7 @@ class SSD(nn.Module):
         "proposal_matcher": det_utils.Matcher,
     }
 
+    @_log_api_usage_once
     def __init__(
         self,
         backbone: nn.Module,
@@ -182,7 +183,6 @@ def __init__(
         positive_fraction: float = 0.25,
     ):
         super().__init__()
-        _log_api_usage_once(self)
 
         self.backbone = backbone
 
diff --git a/torchvision/models/detection/ssdlite.py b/torchvision/models/detection/ssdlite.py
index d35dfbf78d3..b44fa223ccc 100644
--- a/torchvision/models/detection/ssdlite.py
+++ b/torchvision/models/detection/ssdlite.py
@@ -111,6 +111,7 @@ def __init__(self, in_channels: List[int], num_anchors: List[int], norm_layer: C
 
 
 class SSDLiteFeatureExtractorMobileNet(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         backbone: nn.Module,
@@ -120,7 +121,6 @@ def __init__(
         min_depth: int = 16,
     ):
         super().__init__()
-        _log_api_usage_once(self)
 
         assert not backbone[c4_pos].use_res_connect
         self.features = nn.Sequential(
diff --git a/torchvision/models/efficientnet.py b/torchvision/models/efficientnet.py
index 6837018c09e..e1264a93d81 100644
--- a/torchvision/models/efficientnet.py
+++ b/torchvision/models/efficientnet.py
@@ -148,6 +148,7 @@ def forward(self, input: Tensor) -> Tensor:
 
 
 class EfficientNet(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         inverted_residual_setting: List[MBConvConfig],
@@ -170,7 +171,6 @@ def __init__(
             norm_layer (Optional[Callable[..., nn.Module]]): Module specifying the normalization layer to use
         """
         super().__init__()
-        _log_api_usage_once(self)
 
         if not inverted_residual_setting:
             raise ValueError("The inverted_residual_setting should not be empty")
diff --git a/torchvision/models/googlenet.py b/torchvision/models/googlenet.py
index 5e6375c1191..ebba69d2de7 100644
--- a/torchvision/models/googlenet.py
+++ b/torchvision/models/googlenet.py
@@ -28,6 +28,7 @@
 class GoogLeNet(nn.Module):
     __constants__ = ["aux_logits", "transform_input"]
 
+    @_log_api_usage_once
     def __init__(
         self,
         num_classes: int = 1000,
@@ -39,7 +40,6 @@ def __init__(
         dropout_aux: float = 0.7,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         if blocks is None:
             blocks = [BasicConv2d, Inception, InceptionAux]
         if init_weights is None:
diff --git a/torchvision/models/inception.py b/torchvision/models/inception.py
index 322c2370bdd..e48fe627302 100644
--- a/torchvision/models/inception.py
+++ b/torchvision/models/inception.py
@@ -27,6 +27,7 @@
 
 
 class Inception3(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         num_classes: int = 1000,
@@ -37,7 +38,6 @@ def __init__(
         dropout: float = 0.5,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         if inception_blocks is None:
             inception_blocks = [BasicConv2d, InceptionA, InceptionB, InceptionC, InceptionD, InceptionE, InceptionAux]
         if init_weights is None:
diff --git a/torchvision/models/mnasnet.py b/torchvision/models/mnasnet.py
index 5eb27904f90..58de872c2e7 100644
--- a/torchvision/models/mnasnet.py
+++ b/torchvision/models/mnasnet.py
@@ -96,9 +96,9 @@ class MNASNet(torch.nn.Module):
     # Version 2 adds depth scaling in the initial stages of the network.
     _version = 2
 
+    @_log_api_usage_once
     def __init__(self, alpha: float, num_classes: int = 1000, dropout: float = 0.2) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         assert alpha > 0.0
         self.alpha = alpha
         self.num_classes = num_classes
diff --git a/torchvision/models/mobilenetv2.py b/torchvision/models/mobilenetv2.py
index 1a470953df5..38030870f21 100644
--- a/torchvision/models/mobilenetv2.py
+++ b/torchvision/models/mobilenetv2.py
@@ -86,6 +86,7 @@ def forward(self, x: Tensor) -> Tensor:
 
 
 class MobileNetV2(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         num_classes: int = 1000,
@@ -111,7 +112,6 @@ def __init__(
 
         """
         super().__init__()
-        _log_api_usage_once(self)
 
         if block is None:
             block = InvertedResidual
diff --git a/torchvision/models/mobilenetv3.py b/torchvision/models/mobilenetv3.py
index 97239bea8ad..87f0a15efcf 100644
--- a/torchvision/models/mobilenetv3.py
+++ b/torchvision/models/mobilenetv3.py
@@ -129,6 +129,7 @@ def forward(self, input: Tensor) -> Tensor:
 
 
 class MobileNetV3(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         inverted_residual_setting: List[InvertedResidualConfig],
@@ -151,7 +152,6 @@ def __init__(
             dropout (float): The droupout probability
         """
         super().__init__()
-        _log_api_usage_once(self)
 
         if not inverted_residual_setting:
             raise ValueError("The inverted_residual_setting should not be empty")
diff --git a/torchvision/models/regnet.py b/torchvision/models/regnet.py
index 85f53751dd0..c38765c841b 100644
--- a/torchvision/models/regnet.py
+++ b/torchvision/models/regnet.py
@@ -299,6 +299,7 @@ def _adjust_widths_groups_compatibilty(
 
 
 class RegNet(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         block_params: BlockParams,
@@ -310,7 +311,6 @@ def __init__(
         activation: Optional[Callable[..., nn.Module]] = None,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
 
         if stem_type is None:
             stem_type = SimpleStemIN
diff --git a/torchvision/models/resnet.py b/torchvision/models/resnet.py
index b0bb8d13ade..dfa8090a463 100644
--- a/torchvision/models/resnet.py
+++ b/torchvision/models/resnet.py
@@ -162,6 +162,7 @@ def forward(self, x: Tensor) -> Tensor:
 
 
 class ResNet(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         block: Type[Union[BasicBlock, Bottleneck]],
@@ -174,7 +175,6 @@ def __init__(
         norm_layer: Optional[Callable[..., nn.Module]] = None,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         if norm_layer is None:
             norm_layer = nn.BatchNorm2d
         self._norm_layer = norm_layer
diff --git a/torchvision/models/segmentation/_utils.py b/torchvision/models/segmentation/_utils.py
index 0bbea5d3e81..b0c0dd28716 100644
--- a/torchvision/models/segmentation/_utils.py
+++ b/torchvision/models/segmentation/_utils.py
@@ -11,9 +11,9 @@
 class _SimpleSegmentationModel(nn.Module):
     __constants__ = ["aux_classifier"]
 
+    @_log_api_usage_once
     def __init__(self, backbone: nn.Module, classifier: nn.Module, aux_classifier: Optional[nn.Module] = None) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.backbone = backbone
         self.classifier = classifier
         self.aux_classifier = aux_classifier
diff --git a/torchvision/models/segmentation/lraspp.py b/torchvision/models/segmentation/lraspp.py
index f6c2583cac1..9b32264324b 100644
--- a/torchvision/models/segmentation/lraspp.py
+++ b/torchvision/models/segmentation/lraspp.py
@@ -34,11 +34,11 @@ class LRASPP(nn.Module):
         inter_channels (int, optional): the number of channels for intermediate computations.
     """
 
+    @_log_api_usage_once
     def __init__(
         self, backbone: nn.Module, low_channels: int, high_channels: int, num_classes: int, inter_channels: int = 128
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.backbone = backbone
         self.classifier = LRASPPHead(low_channels, high_channels, num_classes, inter_channels)
 
diff --git a/torchvision/models/shufflenetv2.py b/torchvision/models/shufflenetv2.py
index f3758c54aaf..166eaddec06 100644
--- a/torchvision/models/shufflenetv2.py
+++ b/torchvision/models/shufflenetv2.py
@@ -92,6 +92,7 @@ def forward(self, x: Tensor) -> Tensor:
 
 
 class ShuffleNetV2(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         stages_repeats: List[int],
@@ -100,7 +101,6 @@ def __init__(
         inverted_residual: Callable[..., nn.Module] = InvertedResidual,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
 
         if len(stages_repeats) != 3:
             raise ValueError("expected stages_repeats as list of 3 positive ints")
diff --git a/torchvision/models/squeezenet.py b/torchvision/models/squeezenet.py
index 2c1a30f225d..32be85e5020 100644
--- a/torchvision/models/squeezenet.py
+++ b/torchvision/models/squeezenet.py
@@ -34,9 +34,9 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 
 
 class SqueezeNet(nn.Module):
+    @_log_api_usage_once
     def __init__(self, version: str = "1_0", num_classes: int = 1000, dropout: float = 0.5) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.num_classes = num_classes
         if version == "1_0":
             self.features = nn.Sequential(
diff --git a/torchvision/models/vgg.py b/torchvision/models/vgg.py
index e31fc542ca6..f9cb78fc71f 100644
--- a/torchvision/models/vgg.py
+++ b/torchvision/models/vgg.py
@@ -33,11 +33,11 @@
 
 
 class VGG(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self, features: nn.Module, num_classes: int = 1000, init_weights: bool = True, dropout: float = 0.5
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.features = features
         self.avgpool = nn.AdaptiveAvgPool2d((7, 7))
         self.classifier = nn.Sequential(
diff --git a/torchvision/models/video/resnet.py b/torchvision/models/video/resnet.py
index f6899cfafeb..8ca4cd6825a 100644
--- a/torchvision/models/video/resnet.py
+++ b/torchvision/models/video/resnet.py
@@ -188,6 +188,7 @@ def __init__(self) -> None:
 
 
 class VideoResNet(nn.Module):
+    @_log_api_usage_once
     def __init__(
         self,
         block: Type[Union[BasicBlock, Bottleneck]],
@@ -209,7 +210,6 @@ def __init__(
             zero_init_residual (bool, optional): Zero init bottleneck residual BN. Defaults to False.
         """
         super().__init__()
-        _log_api_usage_once(self)
         self.inplanes = 64
 
         self.stem = stem()
diff --git a/torchvision/ops/boxes.py b/torchvision/ops/boxes.py
index 10a03a907e8..3bf61c02085 100644
--- a/torchvision/ops/boxes.py
+++ b/torchvision/ops/boxes.py
@@ -9,6 +9,7 @@
 from ._box_convert import _box_cxcywh_to_xyxy, _box_xyxy_to_cxcywh, _box_xywh_to_xyxy, _box_xyxy_to_xywh
 
 
+@_log_api_usage_once
 def nms(boxes: Tensor, scores: Tensor, iou_threshold: float) -> Tensor:
     """
     Performs non-maximum suppression (NMS) on the boxes according
@@ -34,11 +35,11 @@ def nms(boxes: Tensor, scores: Tensor, iou_threshold: float) -> Tensor:
         Tensor: int64 tensor with the indices of the elements that have been kept
         by NMS, sorted in decreasing order of scores
     """
-    _log_api_usage_once("torchvision.ops.nms")
     _assert_has_ops()
     return torch.ops.torchvision.nms(boxes, scores, iou_threshold)
 
 
+@_log_api_usage_once
 def batched_nms(
     boxes: Tensor,
     scores: Tensor,
@@ -63,7 +64,6 @@ def batched_nms(
         Tensor: int64 tensor with the indices of the elements that have been kept by NMS, sorted
         in decreasing order of scores
     """
-    _log_api_usage_once("torchvision.ops.batched_nms")
     # Benchmarks that drove the following thresholds are at
     # https://github.com/pytorch/vision/issues/1311#issuecomment-781329339
     # Ideally for GPU we'd use a higher threshold
@@ -110,6 +110,7 @@ def _batched_nms_vanilla(
     return keep_indices[scores[keep_indices].sort(descending=True)[1]]
 
 
+@_log_api_usage_once
 def remove_small_boxes(boxes: Tensor, min_size: float) -> Tensor:
     """
     Remove boxes which contains at least one side smaller than min_size.
@@ -123,13 +124,13 @@ def remove_small_boxes(boxes: Tensor, min_size: float) -> Tensor:
         Tensor[K]: indices of the boxes that have both sides
         larger than min_size
     """
-    _log_api_usage_once("torchvision.ops.remove_small_boxes")
     ws, hs = boxes[:, 2] - boxes[:, 0], boxes[:, 3] - boxes[:, 1]
     keep = (ws >= min_size) & (hs >= min_size)
     keep = torch.where(keep)[0]
     return keep
 
 
+@_log_api_usage_once
 def clip_boxes_to_image(boxes: Tensor, size: Tuple[int, int]) -> Tensor:
     """
     Clip boxes so that they lie inside an image of size `size`.
@@ -142,7 +143,6 @@ def clip_boxes_to_image(boxes: Tensor, size: Tuple[int, int]) -> Tensor:
     Returns:
         Tensor[N, 4]: clipped boxes
     """
-    _log_api_usage_once("torchvision.ops.clip_boxes_to_image")
     dim = boxes.dim()
     boxes_x = boxes[..., 0::2]
     boxes_y = boxes[..., 1::2]
@@ -161,6 +161,7 @@ def clip_boxes_to_image(boxes: Tensor, size: Tuple[int, int]) -> Tensor:
     return clipped_boxes.reshape(boxes.shape)
 
 
+@_log_api_usage_once
 def box_convert(boxes: Tensor, in_fmt: str, out_fmt: str) -> Tensor:
     """
     Converts boxes from given in_fmt to out_fmt.
@@ -183,7 +184,6 @@ def box_convert(boxes: Tensor, in_fmt: str, out_fmt: str) -> Tensor:
         Tensor[N, 4]: Boxes into converted format.
     """
 
-    _log_api_usage_once("torchvision.ops.box_convert")
     allowed_fmts = ("xyxy", "xywh", "cxcywh")
     if in_fmt not in allowed_fmts or out_fmt not in allowed_fmts:
         raise ValueError("Unsupported Bounding Box Conversions for given in_fmt and out_fmt")
@@ -220,6 +220,7 @@ def _upcast(t: Tensor) -> Tensor:
         return t if t.dtype in (torch.int32, torch.int64) else t.int()
 
 
+@_log_api_usage_once
 def box_area(boxes: Tensor) -> Tensor:
     """
     Computes the area of a set of bounding boxes, which are specified by their
@@ -233,7 +234,6 @@ def box_area(boxes: Tensor) -> Tensor:
     Returns:
         Tensor[N]: the area for each box
     """
-    _log_api_usage_once("torchvision.ops.box_area")
     boxes = _upcast(boxes)
     return (boxes[:, 2] - boxes[:, 0]) * (boxes[:, 3] - boxes[:, 1])
 
@@ -255,6 +255,7 @@ def _box_inter_union(boxes1: Tensor, boxes2: Tensor) -> Tuple[Tensor, Tensor]:
     return inter, union
 
 
+@_log_api_usage_once
 def box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
     """
     Return intersection-over-union (Jaccard index) between two sets of boxes.
@@ -269,13 +270,13 @@ def box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
     Returns:
         Tensor[N, M]: the NxM matrix containing the pairwise IoU values for every element in boxes1 and boxes2
     """
-    _log_api_usage_once("torchvision.ops.box_iou")
     inter, union = _box_inter_union(boxes1, boxes2)
     iou = inter / union
     return iou
 
 
 # Implementation adapted from https://github.com/facebookresearch/detr/blob/master/util/box_ops.py
+@_log_api_usage_once
 def generalized_box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
     """
     Return generalized intersection-over-union (Jaccard index) between two sets of boxes.
@@ -292,7 +293,6 @@ def generalized_box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
         for every element in boxes1 and boxes2
     """
 
-    _log_api_usage_once("torchvision.ops.generalized_box_iou")
     # degenerate boxes gives inf / nan results
     # so do an early check
     assert (boxes1[:, 2:] >= boxes1[:, :2]).all()
@@ -310,6 +310,7 @@ def generalized_box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
     return iou - (areai - union) / areai
 
 
+@_log_api_usage_once
 def masks_to_boxes(masks: torch.Tensor) -> torch.Tensor:
     """
     Compute the bounding boxes around the provided masks.
@@ -324,7 +325,6 @@ def masks_to_boxes(masks: torch.Tensor) -> torch.Tensor:
     Returns:
         Tensor[N, 4]: bounding boxes
     """
-    _log_api_usage_once("torchvision.ops.masks_to_boxes")
     if masks.numel() == 0:
         return torch.zeros((0, 4), device=masks.device, dtype=torch.float)
 
diff --git a/torchvision/ops/deform_conv.py b/torchvision/ops/deform_conv.py
index 59b1cec049e..55b7ebba075 100644
--- a/torchvision/ops/deform_conv.py
+++ b/torchvision/ops/deform_conv.py
@@ -11,6 +11,7 @@
 from ..utils import _log_api_usage_once
 
 
+@_log_api_usage_once
 def deform_conv2d(
     input: Tensor,
     offset: Tensor,
@@ -61,7 +62,6 @@ def deform_conv2d(
         >>>  torch.Size([4, 5, 8, 8])
     """
 
-    _log_api_usage_once("torchvision.ops.deform_conv2d")
     _assert_has_ops()
     out_channels = weight.shape[0]
 
diff --git a/torchvision/ops/feature_pyramid_network.py b/torchvision/ops/feature_pyramid_network.py
index 93caa47d04b..71a371d646f 100644
--- a/torchvision/ops/feature_pyramid_network.py
+++ b/torchvision/ops/feature_pyramid_network.py
@@ -70,6 +70,7 @@ class FeaturePyramidNetwork(nn.Module):
 
     """
 
+    @_log_api_usage_once
     def __init__(
         self,
         in_channels_list: List[int],
@@ -77,7 +78,6 @@ def __init__(
         extra_blocks: Optional[ExtraFPNBlock] = None,
     ):
         super().__init__()
-        _log_api_usage_once(self)
         self.inner_blocks = nn.ModuleList()
         self.layer_blocks = nn.ModuleList()
         for in_channels in in_channels_list:
diff --git a/torchvision/ops/focal_loss.py b/torchvision/ops/focal_loss.py
index 1a149ed4120..852ec218539 100644
--- a/torchvision/ops/focal_loss.py
+++ b/torchvision/ops/focal_loss.py
@@ -4,6 +4,7 @@
 from ..utils import _log_api_usage_once
 
 
+@_log_api_usage_once
 def sigmoid_focal_loss(
     inputs: torch.Tensor,
     targets: torch.Tensor,
@@ -32,7 +33,6 @@ def sigmoid_focal_loss(
     Returns:
         Loss tensor with the reduction option applied.
     """
-    _log_api_usage_once("torchvision.ops.sigmoid_focal_loss")
     p = torch.sigmoid(inputs)
     ce_loss = F.binary_cross_entropy_with_logits(inputs, targets, reduction="none")
     p_t = p * targets + (1 - p) * (1 - targets)
diff --git a/torchvision/ops/misc.py b/torchvision/ops/misc.py
index fac9a3570d6..17c042c8914 100644
--- a/torchvision/ops/misc.py
+++ b/torchvision/ops/misc.py
@@ -50,6 +50,7 @@ class FrozenBatchNorm2d(torch.nn.Module):
         eps (float): a value added to the denominator for numerical stability. Default: 1e-5
     """
 
+    @_log_api_usage_once
     def __init__(
         self,
         num_features: int,
@@ -61,7 +62,6 @@ def __init__(
             warnings.warn("`n` argument is deprecated and has been renamed `num_features`", DeprecationWarning)
             num_features = n
         super().__init__()
-        _log_api_usage_once(self)
         self.eps = eps
         self.register_buffer("weight", torch.ones(num_features))
         self.register_buffer("bias", torch.zeros(num_features))
@@ -119,6 +119,7 @@ class ConvNormActivation(torch.nn.Sequential):
 
     """
 
+    @_log_api_usage_once
     def __init__(
         self,
         in_channels: int,
@@ -151,7 +152,6 @@ def __init__(
         if activation_layer is not None:
             layers.append(activation_layer(inplace=inplace))
         super().__init__(*layers)
-        _log_api_usage_once(self)
         self.out_channels = out_channels
 
 
@@ -167,6 +167,7 @@ class SqueezeExcitation(torch.nn.Module):
         scale_activation (Callable[..., torch.nn.Module]): ``sigma`` activation. Default: ``torch.nn.Sigmoid``
     """
 
+    @_log_api_usage_once
     def __init__(
         self,
         input_channels: int,
@@ -175,7 +176,6 @@ def __init__(
         scale_activation: Callable[..., torch.nn.Module] = torch.nn.Sigmoid,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
         self.avgpool = torch.nn.AdaptiveAvgPool2d(1)
         self.fc1 = torch.nn.Conv2d(input_channels, squeeze_channels, 1)
         self.fc2 = torch.nn.Conv2d(squeeze_channels, input_channels, 1)
diff --git a/torchvision/ops/poolers.py b/torchvision/ops/poolers.py
index a0cd238dc75..e83160d1c38 100644
--- a/torchvision/ops/poolers.py
+++ b/torchvision/ops/poolers.py
@@ -144,6 +144,7 @@ class MultiScaleRoIAlign(nn.Module):
 
     __annotations__ = {"scales": Optional[List[float]], "map_levels": Optional[LevelMapper]}
 
+    @_log_api_usage_once
     def __init__(
         self,
         featmap_names: List[str],
@@ -154,7 +155,6 @@ def __init__(
         canonical_level: int = 4,
     ):
         super().__init__()
-        _log_api_usage_once(self)
         if isinstance(output_size, int):
             output_size = (output_size, output_size)
         self.featmap_names = featmap_names
diff --git a/torchvision/ops/ps_roi_align.py b/torchvision/ops/ps_roi_align.py
index 4ed4ead89ff..ed25254f06d 100644
--- a/torchvision/ops/ps_roi_align.py
+++ b/torchvision/ops/ps_roi_align.py
@@ -7,6 +7,7 @@
 from ._utils import convert_boxes_to_roi_format, check_roi_boxes_shape
 
 
+@_log_api_usage_once
 def ps_roi_align(
     input: Tensor,
     boxes: Tensor,
@@ -43,7 +44,6 @@ def ps_roi_align(
     Returns:
         Tensor[K, C / (output_size[0] * output_size[1]), output_size[0], output_size[1]]: The pooled RoIs
     """
-    _log_api_usage_once("torchvision.ops.ps_roi_align")
     _assert_has_ops()
     check_roi_boxes_shape(boxes)
     rois = boxes
diff --git a/torchvision/ops/ps_roi_pool.py b/torchvision/ops/ps_roi_pool.py
index 6bab125f04f..c6d1701a917 100644
--- a/torchvision/ops/ps_roi_pool.py
+++ b/torchvision/ops/ps_roi_pool.py
@@ -7,6 +7,7 @@
 from ._utils import convert_boxes_to_roi_format, check_roi_boxes_shape
 
 
+@_log_api_usage_once
 def ps_roi_pool(
     input: Tensor,
     boxes: Tensor,
@@ -37,7 +38,6 @@ def ps_roi_pool(
     Returns:
         Tensor[K, C / (output_size[0] * output_size[1]), output_size[0], output_size[1]]: The pooled RoIs.
     """
-    _log_api_usage_once("torchvision.ops.ps_roi_pool")
     _assert_has_ops()
     check_roi_boxes_shape(boxes)
     rois = boxes
diff --git a/torchvision/ops/roi_align.py b/torchvision/ops/roi_align.py
index 3f80383855b..c5c48a216bc 100644
--- a/torchvision/ops/roi_align.py
+++ b/torchvision/ops/roi_align.py
@@ -10,6 +10,7 @@
 from ._utils import convert_boxes_to_roi_format, check_roi_boxes_shape
 
 
+@_log_api_usage_once
 def roi_align(
     input: Tensor,
     boxes: Union[Tensor, List[Tensor]],
@@ -50,7 +51,6 @@ def roi_align(
     Returns:
         Tensor[K, C, output_size[0], output_size[1]]: The pooled RoIs.
     """
-    _log_api_usage_once("torchvision.ops.roi_align")
     _assert_has_ops()
     check_roi_boxes_shape(boxes)
     rois = boxes
diff --git a/torchvision/ops/roi_pool.py b/torchvision/ops/roi_pool.py
index deef590c953..86134da64dd 100644
--- a/torchvision/ops/roi_pool.py
+++ b/torchvision/ops/roi_pool.py
@@ -10,6 +10,7 @@
 from ._utils import convert_boxes_to_roi_format, check_roi_boxes_shape
 
 
+@_log_api_usage_once
 def roi_pool(
     input: Tensor,
     boxes: Union[Tensor, List[Tensor]],
@@ -39,7 +40,6 @@ def roi_pool(
     Returns:
         Tensor[K, C, output_size[0], output_size[1]]: The pooled RoIs.
     """
-    _log_api_usage_once("torchvision.ops.roi_pool")
     _assert_has_ops()
     check_roi_boxes_shape(boxes)
     rois = boxes
diff --git a/torchvision/ops/stochastic_depth.py b/torchvision/ops/stochastic_depth.py
index b2a0aec233d..999b3c23c08 100644
--- a/torchvision/ops/stochastic_depth.py
+++ b/torchvision/ops/stochastic_depth.py
@@ -5,6 +5,7 @@
 from ..utils import _log_api_usage_once
 
 
+@_log_api_usage_once
 def stochastic_depth(input: Tensor, p: float, mode: str, training: bool = True) -> Tensor:
     """
     Implements the Stochastic Depth from `"Deep Networks with Stochastic Depth"
@@ -23,7 +24,6 @@ def stochastic_depth(input: Tensor, p: float, mode: str, training: bool = True)
     Returns:
         Tensor[N, ...]: The randomly zeroed tensor.
     """
-    _log_api_usage_once("torchvision.ops.stochastic_depth")
     if p < 0.0 or p > 1.0:
         raise ValueError(f"drop probability has to be between 0 and 1, but got {p}")
     if mode not in ["batch", "row"]:
diff --git a/torchvision/utils.py b/torchvision/utils.py
index b11f4ebeecf..bcad3066a49 100644
--- a/torchvision/utils.py
+++ b/torchvision/utils.py
@@ -1,7 +1,8 @@
 import math
 import pathlib
 import warnings
-from typing import Union, Optional, List, Tuple, BinaryIO, no_type_check
+from functools import wraps
+from typing import Union, Optional, List, Tuple, BinaryIO
 
 import numpy as np
 import torch
@@ -375,13 +376,16 @@ def _generate_color_palette(num_masks: int):
     return [tuple((i * palette) % 255) for i in range(num_masks)]
 
 
-@no_type_check
-def _log_api_usage_once(obj: str) -> None:  # type: ignore
-    if torch.jit.is_scripting() or torch.jit.is_tracing():
-        return
-    # NOTE: obj can be an object as well, but mocking it here to be
-    # only a string to appease torchscript
-    if isinstance(obj, str):
-        torch._C._log_api_usage_once(obj)
-    else:
-        torch._C._log_api_usage_once(f"{obj.__module__}.{obj.__class__.__name__}")
+def _log_api_usage_once(f):
+    @wraps(f)
+    def wrapper(*args, **kwargs):
+        event = f.__module__
+        if f.__name__.endswith("__init__"):
+            # inside module instantiation
+            event += args[0].__class__.__name__
+        else:
+            event += f.__name__
+        torch._C._log_api_usage_once(event)
+        return f(*args, **kwargs)
+
+    return wrapper