Fix some tests

Author: NicolasHug

test/test_transforms_v2.py

@@ -492,6 +492,7 @@ def adapt_fill(value, *, dtype):
 def reference_affine_bounding_boxes_helper(bounding_boxes, *, affine_matrix, new_canvas_size=None, clamp=True):
     format = bounding_boxes.format
     canvas_size = new_canvas_size or bounding_boxes.canvas_size
+    clamping_mode = bounding_boxes.clamping_mode
 
     def affine_bounding_boxes(bounding_boxes):
         dtype = bounding_boxes.dtype
@@ -535,6 +536,7 @@ def affine_bounding_boxes(bounding_boxes):
                 output,
                 format=format,
                 canvas_size=canvas_size,
+                clamping_mode=clamping_mode,
             )
         else:
             # We leave the bounding box as float64 so the caller gets the full precision to perform any additional
@@ -557,6 +559,7 @@ def reference_affine_rotated_bounding_boxes_helper(
 ):
     format = bounding_boxes.format
     canvas_size = new_canvas_size or bounding_boxes.canvas_size
+    clamping_mode = bounding_boxes.clamping_mode
 
     def affine_rotated_bounding_boxes(bounding_boxes):
         dtype = bounding_boxes.dtype
@@ -618,6 +621,7 @@ def affine_rotated_bounding_boxes(bounding_boxes):
                 output.to(dtype=dtype, device=device),
                 format=format,
                 canvas_size=canvas_size,
+                clamping_mode=clamping_mode,
             )
             if clamp
             else output.to(dtype=output.dtype, device=device)
@@ -831,7 +835,6 @@ def test_functional(self, size, make_input):
             (F.resize_image, torch.Tensor),
             (F._geometry._resize_image_pil, PIL.Image.Image),
             (F.resize_image, tv_tensors.Image),
-            (F.resize_bounding_boxes, tv_tensors.BoundingBoxes),
             (F.resize_mask, tv_tensors.Mask),
             (F.resize_video, tv_tensors.Video),
             (F.resize_keypoints, tv_tensors.KeyPoints),
@@ -3289,7 +3292,6 @@ def test_functional(self, make_input):
             (F.elastic_image, torch.Tensor),
             (F._geometry._elastic_image_pil, PIL.Image.Image),
             (F.elastic_image, tv_tensors.Image),
-            (F.elastic_bounding_boxes, tv_tensors.BoundingBoxes),
             (F.elastic_mask, tv_tensors.Mask),
             (F.elastic_video, tv_tensors.Video),
             (F.elastic_keypoints, tv_tensors.KeyPoints),
@@ -5126,6 +5128,7 @@ def test_image_functional_correctness(self, coefficients, interpolation, fill):
     def _reference_perspective_bounding_boxes(self, bounding_boxes, *, startpoints, endpoints):
         format = bounding_boxes.format
         canvas_size = bounding_boxes.canvas_size
+        clamping_mode = bounding_boxes.clamping_mode
         dtype = bounding_boxes.dtype
         device = bounding_boxes.device
         is_rotated = tv_tensors.is_rotated_bounding_format(format)
@@ -5226,6 +5229,7 @@ def perspective_bounding_boxes(bounding_boxes):
                 output,
                 format=format,
                 canvas_size=canvas_size,
+                clamping_mode=clamping_mode,
             ).to(dtype=dtype, device=device)
 
         return tv_tensors.BoundingBoxes(

torchvision/transforms/v2/functional/_geometry.py

@@ -9,6 +9,7 @@
 from torch.nn.functional import grid_sample, interpolate, pad as torch_pad
 
 from torchvision import tv_tensors
+from torchvision.tv_tensors._bounding_boxes import CLAMPING_MODE_TYPE 
 from torchvision.transforms import _functional_pil as _FP
 from torchvision.transforms._functional_tensor import _pad_symmetric
 from torchvision.transforms.functional import (
@@ -521,6 +522,7 @@ def resize_bounding_boxes(
     size: Optional[list[int]],
     max_size: Optional[int] = None,
     format: tv_tensors.BoundingBoxFormat = tv_tensors.BoundingBoxFormat.XYXY,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> tuple[torch.Tensor, tuple[int, int]]:
     # We set the default format as `tv_tensors.BoundingBoxFormat.XYXY`
     # to ensure backward compatibility.
@@ -546,7 +548,7 @@ def resize_bounding_boxes(
         transformed_points = xyxyxyxy_boxes.mul(ratios)
         out_bboxes = _parallelogram_to_bounding_boxes(transformed_points)
         out_bboxes = clamp_bounding_boxes(
-            out_bboxes, format=tv_tensors.BoundingBoxFormat.XYXYXYXY, canvas_size=(new_height, new_width)
+            out_bboxes, format=tv_tensors.BoundingBoxFormat.XYXYXYXY, canvas_size=(new_height, new_width), clamping_mode=clamping_mode,
         )
         return (
             convert_bounding_box_format(
@@ -572,7 +574,7 @@ def _resize_bounding_boxes_dispatch(
     inpt: tv_tensors.BoundingBoxes, size: Optional[list[int]], max_size: Optional[int] = None, **kwargs: Any
 ) -> tv_tensors.BoundingBoxes:
     output, canvas_size = resize_bounding_boxes(
-        inpt.as_subclass(torch.Tensor), format=inpt.format, canvas_size=inpt.canvas_size, size=size, max_size=max_size
+        inpt.as_subclass(torch.Tensor), format=inpt.format, canvas_size=inpt.canvas_size, size=size, max_size=max_size, clamping_mode=inpt.clamping_mode
     )
     return tv_tensors.wrap(output, like=inpt, canvas_size=canvas_size)
 
@@ -1098,6 +1100,7 @@ def _affine_bounding_boxes_with_expand(
     shear: list[float],
     center: Optional[list[float]] = None,
     expand: bool = False,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> tuple[torch.Tensor, tuple[int, int]]:
     if bounding_boxes.numel() == 0:
         return bounding_boxes, canvas_size
@@ -1176,7 +1179,7 @@ def _affine_bounding_boxes_with_expand(
         new_width, new_height = _compute_affine_output_size(affine_vector, width, height)
         canvas_size = (new_height, new_width)
 
-    out_bboxes = clamp_bounding_boxes(out_bboxes, format=intermediate_format, canvas_size=canvas_size)
+    out_bboxes = clamp_bounding_boxes(out_bboxes, format=intermediate_format, canvas_size=canvas_size, clamping_mode=clamping_mode)
     out_bboxes = convert_bounding_box_format(
         out_bboxes, old_format=intermediate_format, new_format=format, inplace=True
     ).reshape(original_shape)
@@ -1197,6 +1200,7 @@ def affine_bounding_boxes(
     scale: float,
     shear: list[float],
     center: Optional[list[float]] = None,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> torch.Tensor:
     out_box, _ = _affine_bounding_boxes_with_expand(
         bounding_boxes,
@@ -1208,6 +1212,7 @@ def affine_bounding_boxes(
         shear=shear,
         center=center,
         expand=False,
+        clamping_mode=clamping_mode,
     )
     return out_box
 
@@ -1231,6 +1236,7 @@ def _affine_bounding_boxes_dispatch(
         scale=scale,
         shear=shear,
         center=center,
+        clamping_mode=inpt.clamping_mode,
     )
     return tv_tensors.wrap(output, like=inpt)
 
@@ -1723,6 +1729,7 @@ def pad_bounding_boxes(
     canvas_size: tuple[int, int],
     padding: list[int],
     padding_mode: str = "constant",
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> tuple[torch.Tensor, tuple[int, int]]:
     if padding_mode not in ["constant"]:
         # TODO: add support of other padding modes
@@ -1745,7 +1752,7 @@ def pad_bounding_boxes(
     width += left + right
     canvas_size = (height, width)
 
-    return clamp_bounding_boxes(bounding_boxes, format=format, canvas_size=canvas_size), canvas_size
+    return clamp_bounding_boxes(bounding_boxes, format=format, canvas_size=canvas_size, clamping_mode=clamping_mode), canvas_size
 
 
 @_register_kernel_internal(pad, tv_tensors.BoundingBoxes, tv_tensor_wrapper=False)
@@ -1758,6 +1765,7 @@ def _pad_bounding_boxes_dispatch(
         canvas_size=inpt.canvas_size,
         padding=padding,
         padding_mode=padding_mode,
+        clamping_mode=inpt.clamping_mode
     )
     return tv_tensors.wrap(output, like=inpt, canvas_size=canvas_size)
 
@@ -1836,6 +1844,7 @@ def crop_bounding_boxes(
     left: int,
     height: int,
     width: int,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> tuple[torch.Tensor, tuple[int, int]]:
 
     # Crop or implicit pad if left and/or top have negative values:
@@ -1854,15 +1863,15 @@ def crop_bounding_boxes(
     if format == tv_tensors.BoundingBoxFormat.XYXYXYXY:
         bounding_boxes = _parallelogram_to_bounding_boxes(bounding_boxes)
 
-    return clamp_bounding_boxes(bounding_boxes, format=format, canvas_size=canvas_size), canvas_size
+    return clamp_bounding_boxes(bounding_boxes, format=format, canvas_size=canvas_size, clamping_mode=clamping_mode), canvas_size
 
 
 @_register_kernel_internal(crop, tv_tensors.BoundingBoxes, tv_tensor_wrapper=False)
 def _crop_bounding_boxes_dispatch(
     inpt: tv_tensors.BoundingBoxes, top: int, left: int, height: int, width: int
 ) -> tv_tensors.BoundingBoxes:
     output, canvas_size = crop_bounding_boxes(
-        inpt.as_subclass(torch.Tensor), format=inpt.format, top=top, left=left, height=height, width=width
+        inpt.as_subclass(torch.Tensor), format=inpt.format, top=top, left=left, height=height, width=width, clamping_mode=inpt.clamping_mode
     )
     return tv_tensors.wrap(output, like=inpt, canvas_size=canvas_size)
 
@@ -2066,6 +2075,7 @@ def perspective_bounding_boxes(
     startpoints: Optional[list[list[int]]],
     endpoints: Optional[list[list[int]]],
     coefficients: Optional[list[float]] = None,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> torch.Tensor:
     if bounding_boxes.numel() == 0:
         return bounding_boxes
@@ -2130,7 +2140,7 @@ def perspective_bounding_boxes(
         out_bbox_mins, out_bbox_maxs = torch.aminmax(transformed_points, dim=1)
         out_bboxes = torch.cat([out_bbox_mins, out_bbox_maxs], dim=1)
 
-    out_bboxes = clamp_bounding_boxes(out_bboxes, format=intermediate_format, canvas_size=canvas_size)
+    out_bboxes = clamp_bounding_boxes(out_bboxes, format=intermediate_format, canvas_size=canvas_size, clamping_mode=clamping_mode)
 
     out_bboxes = convert_bounding_box_format(
         out_bboxes, old_format=intermediate_format, new_format=format, inplace=True
@@ -2185,6 +2195,7 @@ def _perspective_bounding_boxes_dispatch(
         startpoints=startpoints,
         endpoints=endpoints,
         coefficients=coefficients,
+        clamping_mode=inpt.clamping_mode,
     )
     return tv_tensors.wrap(output, like=inpt)
 
@@ -2377,6 +2388,7 @@ def elastic_bounding_boxes(
     format: tv_tensors.BoundingBoxFormat,
     canvas_size: tuple[int, int],
     displacement: torch.Tensor,
+    clamping_mode: CLAMPING_MODE_TYPE = "hard",  # TODOBB soft
 ) -> torch.Tensor:
     expected_shape = (1, canvas_size[0], canvas_size[1], 2)
     if not isinstance(displacement, torch.Tensor):
@@ -2432,6 +2444,7 @@ def elastic_bounding_boxes(
         out_bboxes,
         format=intermediate_format,
         canvas_size=canvas_size,
+        clamping_mode=clamping_mode
     )
 
     return convert_bounding_box_format(
@@ -2444,7 +2457,7 @@ def _elastic_bounding_boxes_dispatch(
     inpt: tv_tensors.BoundingBoxes, displacement: torch.Tensor, **kwargs
 ) -> tv_tensors.BoundingBoxes:
     output = elastic_bounding_boxes(
-        inpt.as_subclass(torch.Tensor), format=inpt.format, canvas_size=inpt.canvas_size, displacement=displacement
+        inpt.as_subclass(torch.Tensor), format=inpt.format, canvas_size=inpt.canvas_size, displacement=displacement, clamping_mode=inpt.clamping_mode
     )
     return tv_tensors.wrap(output, like=inpt)