diff --git a/torchvision/models/detection/transform.py b/torchvision/models/detection/transform.py
index 0d95361eedb..5f1dabfaa85 100644
--- a/torchvision/models/detection/transform.py
+++ b/torchvision/models/detection/transform.py
@@ -27,7 +27,7 @@ def _resize_image_and_masks_onnx(image, self_min_size, self_max_size, target):
 
     if "masks" in target:
         mask = target["masks"]
-        mask = F.interpolate(mask[:, None].float(), scale_factor=scale_factor)[:, 0].byte()
+        mask = F.interpolate(mask[:, None].float(), scale_factor=scale_factor, recompute_scale_factor=True)[:, 0].byte()
         target["masks"] = mask
     return image, target
 
@@ -49,7 +49,7 @@ def _resize_image_and_masks(image, self_min_size, self_max_size, target):
 
     if "masks" in target:
         mask = target["masks"]
-        mask = F.interpolate(mask[:, None].float(), scale_factor=scale_factor)[:, 0].byte()
+        mask = F.interpolate(mask[:, None].float(), scale_factor=scale_factor, recompute_scale_factor=True)[:, 0].byte()
         target["masks"] = mask
     return image, target