Merge branch 'main' into v2-vs-v1-fixes

Author: vfdev-5

docs/source/transforms.rst

@@ -228,12 +228,11 @@ Conversion
 
     ToPILImage
     v2.ToPILImage
-    v2.ToImagePIL
     ToTensor
     v2.ToTensor
     PILToTensor
     v2.PILToTensor
-    v2.ToImageTensor
+    v2.ToImage
     ConvertImageDtype
     v2.ConvertImageDtype
     v2.ToDtype

gallery/plot_transforms_v2_e2e.py

@@ -27,7 +27,7 @@ def show(sample):
 
     image, target = sample
     if isinstance(image, PIL.Image.Image):
-        image = F.to_image_tensor(image)
+        image = F.to_image(image)
     image = F.to_dtype(image, torch.uint8, scale=True)
     annotated_image = draw_bounding_boxes(image, target["boxes"], colors="yellow", width=3)
 
@@ -101,7 +101,7 @@ def load_example_coco_detection_dataset(**kwargs):
         transforms.RandomZoomOut(fill={PIL.Image.Image: (123, 117, 104), "others": 0}),
         transforms.RandomIoUCrop(),
         transforms.RandomHorizontalFlip(),
-        transforms.ToImageTensor(),
+        transforms.ToImage(),
         transforms.ConvertImageDtype(torch.float32),
         transforms.SanitizeBoundingBoxes(),
     ]

packaging/pre_build_script.sh

@@ -11,7 +11,7 @@ if [[ "$(uname)" == Darwin ]]; then
   conda install -yq wget
 fi
 
-if [[ "$(uname)" == Darwin || "$OSTYPE" == "msys" ]]; then
+if [[ "$(uname)" == Darwin || "$OSTYPE" == "msys" || "$ARCH" == "aarch64" ]]; then
   # Install libpng from Anaconda (defaults)
   conda install libpng -yq
   conda install -yq ffmpeg=4.2 libjpeg-turbo -c pytorch

references/detection/presets.py

@@ -33,7 +33,7 @@ def __init__(
         transforms = []
         backend = backend.lower()
         if backend == "datapoint":
-            transforms.append(T.ToImageTensor())
+            transforms.append(T.ToImage())
         elif backend == "tensor":
             transforms.append(T.PILToTensor())
         elif backend != "pil":
@@ -71,7 +71,7 @@ def __init__(
 
         if backend == "pil":
             # Note: we could just convert to pure tensors even in v2.
-            transforms += [T.ToImageTensor() if use_v2 else T.PILToTensor()]
+            transforms += [T.ToImage() if use_v2 else T.PILToTensor()]
 
         transforms += [T.ConvertImageDtype(torch.float)]
 
@@ -94,11 +94,11 @@ def __init__(self, backend="pil", use_v2=False):
         backend = backend.lower()
         if backend == "pil":
             # Note: we could just convert to pure tensors even in v2?
-            transforms += [T.ToImageTensor() if use_v2 else T.PILToTensor()]
+            transforms += [T.ToImage() if use_v2 else T.PILToTensor()]
         elif backend == "tensor":
             transforms += [T.PILToTensor()]
         elif backend == "datapoint":
-            transforms += [T.ToImageTensor()]
+            transforms += [T.ToImage()]
         else:
             raise ValueError(f"backend can be 'datapoint', 'tensor' or 'pil', but got {backend}")
 

references/segmentation/presets.py

@@ -32,7 +32,7 @@ def __init__(
         transforms = []
         backend = backend.lower()
         if backend == "datapoint":
-            transforms.append(T.ToImageTensor())
+            transforms.append(T.ToImage())
         elif backend == "tensor":
             transforms.append(T.PILToTensor())
         elif backend != "pil":
@@ -81,7 +81,7 @@ def __init__(
         if backend == "tensor":
             transforms += [T.PILToTensor()]
         elif backend == "datapoint":
-            transforms += [T.ToImageTensor()]
+            transforms += [T.ToImage()]
         elif backend != "pil":
             raise ValueError(f"backend can be 'datapoint', 'tensor' or 'pil', but got {backend}")
 
@@ -92,7 +92,7 @@ def __init__(
 
         if backend == "pil":
             # Note: we could just convert to pure tensors even in v2?
-            transforms += [T.ToImageTensor() if use_v2 else T.PILToTensor()]
+            transforms += [T.ToImage() if use_v2 else T.PILToTensor()]
 
         transforms += [
             T.ConvertImageDtype(torch.float),

test/common_utils.py

@@ -27,7 +27,7 @@
 from torch.testing._comparison import BooleanPair, NonePair, not_close_error_metas, NumberPair, TensorLikePair
 from torchvision import datapoints, io
 from torchvision.transforms._functional_tensor import _max_value as get_max_value
-from torchvision.transforms.v2.functional import to_dtype_image_tensor, to_image_pil, to_image_tensor
+from torchvision.transforms.v2.functional import to_dtype_image, to_image, to_pil_image
 
 
 IN_OSS_CI = any(os.getenv(var) == "true" for var in ["CIRCLECI", "GITHUB_ACTIONS"])
@@ -293,7 +293,7 @@ def __init__(
         **other_parameters,
     ):
         if all(isinstance(input, PIL.Image.Image) for input in [actual, expected]):
-            actual, expected = [to_image_tensor(input) for input in [actual, expected]]
+            actual, expected = [to_image(input) for input in [actual, expected]]
 
         super().__init__(actual, expected, **other_parameters)
         self.mae = mae
@@ -536,7 +536,7 @@ def make_image_tensor(*args, **kwargs):
 
 
 def make_image_pil(*args, **kwargs):
-    return to_image_pil(make_image(*args, **kwargs))
+    return to_pil_image(make_image(*args, **kwargs))
 
 
 def make_image_loader(
@@ -609,12 +609,12 @@ def fn(shape, dtype, device, memory_format):
             )
         )
 
-        image_tensor = to_image_tensor(image_pil)
+        image_tensor = to_image(image_pil)
         if memory_format == torch.contiguous_format:
             image_tensor = image_tensor.to(device=device, memory_format=memory_format, copy=True)
         else:
             image_tensor = image_tensor.to(device=device)
-        image_tensor = to_dtype_image_tensor(image_tensor, dtype=dtype, scale=True)
+        image_tensor = to_dtype_image(image_tensor, dtype=dtype, scale=True)
 
         return datapoints.Image(image_tensor)
 

test/test_prototype_transforms.py

@@ -17,7 +17,7 @@
 
 from torchvision.datapoints import BoundingBoxes, BoundingBoxFormat, Image, Mask, Video
 from torchvision.prototype import datapoints, transforms
-from torchvision.transforms.v2.functional import clamp_bounding_boxes, InterpolationMode, pil_to_tensor, to_image_pil
+from torchvision.transforms.v2.functional import clamp_bounding_boxes, InterpolationMode, pil_to_tensor, to_pil_image
 from torchvision.transforms.v2.utils import check_type, is_simple_tensor
 
 BATCH_EXTRA_DIMS = [extra_dims for extra_dims in DEFAULT_EXTRA_DIMS if extra_dims]
@@ -387,7 +387,7 @@ def make_datapoints():
         size = (600, 800)
         num_objects = 22
 
-        pil_image = to_image_pil(make_image(size=size, color_space="RGB"))
+        pil_image = to_pil_image(make_image(size=size, color_space="RGB"))
         target = {
             "boxes": make_bounding_box(canvas_size=size, format="XYXY", batch_dims=(num_objects,), dtype=torch.float),
             "labels": make_label(extra_dims=(num_objects,), categories=80),

test/test_transforms_v2.py

@@ -666,50 +666,33 @@ def test_check_transformed_types(self, inpt_type, mocker):
                 t(inpt)
 
 
-class TestToImageTensor:
+class TestToImage:
     @pytest.mark.parametrize(
         "inpt_type",
         [torch.Tensor, PIL.Image.Image, datapoints.Image, np.ndarray, datapoints.BoundingBoxes, str, int],
     )
     def test__transform(self, inpt_type, mocker):
         fn = mocker.patch(
-            "torchvision.transforms.v2.functional.to_image_tensor",
+            "torchvision.transforms.v2.functional.to_image",
             return_value=torch.rand(1, 3, 8, 8),
         )
 
         inpt = mocker.MagicMock(spec=inpt_type)
-        transform = transforms.ToImageTensor()
+        transform = transforms.ToImage()
         transform(inpt)
         if inpt_type in (datapoints.BoundingBoxes, datapoints.Image, str, int):
             assert fn.call_count == 0
         else:
             fn.assert_called_once_with(inpt)
 
 
-class TestToImagePIL:
-    @pytest.mark.parametrize(
-        "inpt_type",
-        [torch.Tensor, PIL.Image.Image, datapoints.Image, np.ndarray, datapoints.BoundingBoxes, str, int],
-    )
-    def test__transform(self, inpt_type, mocker):
-        fn = mocker.patch("torchvision.transforms.v2.functional.to_image_pil")
-
-        inpt = mocker.MagicMock(spec=inpt_type)
-        transform = transforms.ToImagePIL()
-        transform(inpt)
-        if inpt_type in (datapoints.BoundingBoxes, PIL.Image.Image, str, int):
-            assert fn.call_count == 0
-        else:
-            fn.assert_called_once_with(inpt, mode=transform.mode)
-
-
 class TestToPILImage:
     @pytest.mark.parametrize(
         "inpt_type",
         [torch.Tensor, PIL.Image.Image, datapoints.Image, np.ndarray, datapoints.BoundingBoxes, str, int],
     )
     def test__transform(self, inpt_type, mocker):
-        fn = mocker.patch("torchvision.transforms.v2.functional.to_image_pil")
+        fn = mocker.patch("torchvision.transforms.v2.functional.to_pil_image")
 
         inpt = mocker.MagicMock(spec=inpt_type)
         transform = transforms.ToPILImage()
@@ -1013,7 +996,7 @@ def test_antialias_warning():
 @pytest.mark.parametrize("image_type", (PIL.Image, torch.Tensor, datapoints.Image))
 @pytest.mark.parametrize("label_type", (torch.Tensor, int))
 @pytest.mark.parametrize("dataset_return_type", (dict, tuple))
-@pytest.mark.parametrize("to_tensor", (transforms.ToTensor, transforms.ToImageTensor))
+@pytest.mark.parametrize("to_tensor", (transforms.ToTensor, transforms.ToImage))
 def test_classif_preset(image_type, label_type, dataset_return_type, to_tensor):
 
     image = datapoints.Image(torch.randint(0, 256, size=(1, 3, 250, 250), dtype=torch.uint8))
@@ -1074,7 +1057,7 @@ def test_classif_preset(image_type, label_type, dataset_return_type, to_tensor):
 
 @pytest.mark.parametrize("image_type", (PIL.Image, torch.Tensor, datapoints.Image))
 @pytest.mark.parametrize("data_augmentation", ("hflip", "lsj", "multiscale", "ssd", "ssdlite"))
-@pytest.mark.parametrize("to_tensor", (transforms.ToTensor, transforms.ToImageTensor))
+@pytest.mark.parametrize("to_tensor", (transforms.ToTensor, transforms.ToImage))
 @pytest.mark.parametrize("sanitize", (True, False))
 def test_detection_preset(image_type, data_augmentation, to_tensor, sanitize):
     torch.manual_seed(0)

test/test_transforms_v2_consistency.py

@@ -30,7 +30,7 @@
 from torchvision.transforms import functional as legacy_F
 from torchvision.transforms.v2 import functional as prototype_F
 from torchvision.transforms.v2._utils import _get_fill
-from torchvision.transforms.v2.functional import to_image_pil
+from torchvision.transforms.v2.functional import to_pil_image
 from torchvision.transforms.v2.utils import query_size
 
 DEFAULT_MAKE_IMAGES_KWARGS = dict(color_spaces=["RGB"], extra_dims=[(4,)])
@@ -630,7 +630,7 @@ def check_call_consistency(
         )
 
         if image.ndim == 3 and supports_pil:
-            image_pil = to_image_pil(image)
+            image_pil = to_pil_image(image)
 
             try:
                 torch.manual_seed(0)
@@ -869,7 +869,7 @@ def test_pil_to_tensor(self):
         legacy_transform = legacy_transforms.PILToTensor()
 
         for image in make_images(extra_dims=[()]):
-            image_pil = to_image_pil(image)
+            image_pil = to_pil_image(image)
 
             assert_equal(prototype_transform(image_pil), legacy_transform(image_pil))
 
@@ -879,7 +879,7 @@ def test_to_tensor(self):
         legacy_transform = legacy_transforms.ToTensor()
 
         for image in make_images(extra_dims=[()]):
-            image_pil = to_image_pil(image)
+            image_pil = to_pil_image(image)
             image_numpy = np.array(image_pil)
 
             assert_equal(prototype_transform(image_pil), legacy_transform(image_pil))
@@ -1182,7 +1182,7 @@ def make_datapoints(self, with_mask=True):
         def make_label(extra_dims, categories):
             return torch.randint(categories, extra_dims, dtype=torch.int64)
 
-        pil_image = to_image_pil(make_image(size=size, color_space="RGB"))
+        pil_image = to_pil_image(make_image(size=size, color_space="RGB"))
         target = {
             "boxes": make_bounding_box(canvas_size=size, format="XYXY", batch_dims=(num_objects,), dtype=torch.float),
             "labels": make_label(extra_dims=(num_objects,), categories=80),
@@ -1286,7 +1286,7 @@ def make_datapoints(self, supports_pil=True, image_dtype=torch.uint8):
 
         conv_fns = []
         if supports_pil:
-            conv_fns.append(to_image_pil)
+            conv_fns.append(to_pil_image)
         conv_fns.extend([torch.Tensor, lambda x: x])
 
         for conv_fn in conv_fns:
@@ -1295,8 +1295,8 @@ def make_datapoints(self, supports_pil=True, image_dtype=torch.uint8):
 
             dp = (conv_fn(datapoint_image), datapoint_mask)
             dp_ref = (
-                to_image_pil(datapoint_image) if supports_pil else datapoint_image.as_subclass(torch.Tensor),
-                to_image_pil(datapoint_mask),
+                to_pil_image(datapoint_image) if supports_pil else datapoint_image.as_subclass(torch.Tensor),
+                to_pil_image(datapoint_mask),
             )
 
             yield dp, dp_ref