Recoded gaussian blur without using to_tensor (#5503)

torchvision/prototype/transforms/functional/_misc.py

@@ -3,7 +3,7 @@ from typing import Optional, List
 import PIL.Image
 import torch
 from torchvision.transforms import functional_tensor as _FT
-from torchvision.transforms.functional import to_tensor, to_pil_image
+from torchvision.transforms.functional import pil_to_tensor, to_pil_image
 
 
 normalize_image_tensor = _FT.normalize
@@ -39,4 +39,6 @@ def gaussian_blur_image_tensor(
 
 
 def gaussian_blur_image_pil(img: PIL.Image, kernel_size: List[int], sigma: Optional[List[float]] = None) -> PIL.Image:
-    return to_pil_image(gaussian_blur_image_tensor(to_tensor(img), kernel_size=kernel_size, sigma=sigma))
+    t_img = pil_to_tensor(img)
+    output = gaussian_blur_image_tensor(t_img, kernel_size=kernel_size, sigma=sigma)
+    return to_pil_image(output, mode=img.mode)

torchvision/transforms/functional.py

@@ -403,11 +403,8 @@ def resize(
             mode).
         antialias (bool, optional): antialias flag. If ``img`` is PIL Image, the flag is ignored and anti-alias
             is always used. If ``img`` is Tensor, the flag is False by default and can be set to True for
-            ``InterpolationMode.BILINEAR`` only mode. This can help making the output for PIL images and tensors
-            closer.
-
-            .. warning::
-                There is no autodiff support for ``antialias=True`` option with input ``img`` as Tensor.
+            ``InterpolationMode.BILINEAR`` and ``InterpolationMode.BICUBIC`` modes.
+            This can help making the output for PIL images and tensors closer.
 
     Returns:
         PIL Image or Tensor: Resized image.
@@ -1338,12 +1335,12 @@ def gaussian_blur(img: Tensor, kernel_size: List[int], sigma: Optional[List[floa
         if not F_pil._is_pil_image(img):
             raise TypeError(f"img should be PIL Image or Tensor. Got {type(img)}")
 
-        t_img = to_tensor(img)
+        t_img = pil_to_tensor(img)
 
     output = F_t.gaussian_blur(t_img, kernel_size, sigma)
 
     if not isinstance(img, torch.Tensor):
-        output = to_pil_image(output)
+        output = to_pil_image(output, mode=img.mode)
     return output