Skip to content

add tests for F.pad_bounding_box #6038

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
May 26, 2022
Merged

add tests for F.pad_bounding_box #6038

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
ca7d822
add tests for F.pad_bounding_box
pmeier May 17, 2022
ee388be
Merge branch 'main' into test-pad-bbox
vfdev-5 May 23, 2022
d3ae52a
Merge branch 'main' of github.com:pytorch/vision into test-pad-bbox
vfdev-5 May 25, 2022
5945b41
Added correctness tests for pad and reimplemented bbox op to keep dtype
vfdev-5 May 26, 2022
8355db2
Merge branch 'main' into test-pad-bbox
vfdev-5 May 26, 2022
a7fd05c
Update _geometry.py
vfdev-5 May 26, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
82 changes: 68 additions & 14 deletions test/test_prototype_transforms_functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -382,6 +382,15 @@ def pad_segmentation_mask():
yield SampleInput(mask, padding=padding, padding_mode=padding_mode)


@register_kernel_info_from_sample_inputs_fn
def pad_bounding_box():
for bounding_box, padding in itertools.product(
make_bounding_boxes(),
[[1], [1, 1], [1, 1, 2, 2]],
):
yield SampleInput(bounding_box, padding=padding, format=bounding_box.format)


@register_kernel_info_from_sample_inputs_fn
def perspective_bounding_box():
for bounding_box, perspective_coeffs in itertools.product(
Expand Down Expand Up @@ -1103,22 +1112,67 @@ def test_correctness_pad_segmentation_mask_on_fixed_input(device):
torch.testing.assert_close(out_mask, expected_mask)


def _parse_padding(padding):
if isinstance(padding, int):
return [padding] * 4
if isinstance(padding, list):
if len(padding) == 1:
return padding * 4
if len(padding) == 2:
return padding * 2 # [left, up, right, down]

return padding


@pytest.mark.parametrize("device", cpu_and_gpu())
@pytest.mark.parametrize("padding", [[1], [1, 1], [1, 1, 2, 2]])
def test_correctness_pad_bounding_box(device, padding):
def _compute_expected_bbox(bbox, padding_):
pad_left, pad_up, _, _ = _parse_padding(padding_)

bbox_format = bbox.format
bbox_dtype = bbox.dtype
bbox = convert_bounding_box_format(bbox, old_format=bbox_format, new_format=features.BoundingBoxFormat.XYXY)

bbox[0::2] += pad_left
bbox[1::2] += pad_up

bbox = convert_bounding_box_format(
bbox, old_format=features.BoundingBoxFormat.XYXY, new_format=bbox_format, copy=False
)
if bbox.dtype != bbox_dtype:
# Temporary cast to original dtype
# e.g. float32 -> int
bbox = bbox.to(bbox_dtype)
return bbox
Comment on lines +1140 to +1147
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. There is no need for this if: .to is a no-op if no conversion needs to be performed.
  2. You don't need to use the dtype explicitly. You can pass a tensor to .to in which case it will convert device and dtype to the input.
Suggested change
bbox = convert_bounding_box_format(
bbox, old_format=features.BoundingBoxFormat.XYXY, new_format=bbox_format, copy=False
)
if bbox.dtype != bbox_dtype:
# Temporary cast to original dtype
# e.g. float32 -> int
bbox = bbox.to(bbox_dtype)
return bbox
return convert_bounding_box_format(
bbox, old_format=features.BoundingBoxFormat.XYXY, new_format=bbox_.format, copy=False
).to(bbox_)

Copy link
Collaborator

@vfdev-5 vfdev-5 May 26, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@pmeier actually it does not work with .to between tensors and BoundingBox:

>>> t = torch.tensor([[1.0, 2.0, 3.0, 4.0]])
>>> bbox = features.BoundingBox([[0, 1, 2, 3]], format=features.BoundingBoxFormat.XYXY, image_size=(32, 32))
>>> t.to(bbox)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/vision/torchvision/prototype/features/_feature.py", line 83, in __torch_function__
    return cls.new_like(args[0], output, dtype=output.dtype, device=output.device)
  File "/vision/torchvision/prototype/features/_bounding_box.py", line 54, in new_like
    format=format if format is not None else other.format,
AttributeError: 'Tensor' object has no attribute 'format'

I'll keep my code, especially it is a temporary cast once we fixed convert_bounding_box_format we can remove it quickly

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'll added #6094 to track this.


for bboxes in make_bounding_boxes():
bboxes = bboxes.to(device)
bboxes_format = bboxes.format
bboxes_image_size = bboxes.image_size

output_boxes = F.pad_bounding_box(bboxes, padding, format=bboxes_format)

if bboxes.ndim < 2:
bboxes = [bboxes]

expected_bboxes = []
for bbox in bboxes:
bbox = features.BoundingBox(bbox, format=bboxes_format, image_size=bboxes_image_size)
expected_bboxes.append(_compute_expected_bbox(bbox, padding))

if len(expected_bboxes) > 1:
expected_bboxes = torch.stack(expected_bboxes)
else:
expected_bboxes = expected_bboxes[0]
torch.testing.assert_close(output_boxes, expected_bboxes)


@pytest.mark.parametrize("padding", [[1, 2, 3, 4], [1], 1, [1, 2]])
def test_correctness_pad_segmentation_mask(padding):
def _compute_expected_mask():
def parse_padding():
if isinstance(padding, int):
return [padding] * 4
if isinstance(padding, list):
if len(padding) == 1:
return padding * 4
if len(padding) == 2:
return padding * 2 # [left, up, right, down]

return padding

def _compute_expected_mask(mask, padding_):
h, w = mask.shape[-2], mask.shape[-1]
pad_left, pad_up, pad_right, pad_down = parse_padding()
pad_left, pad_up, pad_right, pad_down = _parse_padding(padding_)

new_h = h + pad_up + pad_down
new_w = w + pad_left + pad_right
Expand All @@ -1132,7 +1186,7 @@ def parse_padding():
for mask in make_segmentation_masks():
out_mask = F.pad_segmentation_mask(mask, padding, "constant")

expected_mask = _compute_expected_mask()
expected_mask = _compute_expected_mask(mask, padding)
torch.testing.assert_close(out_mask, expected_mask)


Expand Down
19 changes: 9 additions & 10 deletions torchvision/prototype/transforms/functional/_geometry.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -415,16 +415,15 @@ def pad_bounding_box(
) -> torch.Tensor:
left, _, top, _ = _FT._parse_pad_padding(padding)

bounding_box = convert_bounding_box_format(
bounding_box, old_format=format, new_format=features.BoundingBoxFormat.XYXY
)

bounding_box[..., 0::2] += left
bounding_box[..., 1::2] += top

return convert_bounding_box_format(
bounding_box, old_format=features.BoundingBoxFormat.XYXY, new_format=format, copy=False
)
bounding_box = bounding_box.clone()

# this works without conversion since padding only affects xy coordinates
bounding_box[..., 0] += left
bounding_box[..., 1] += top
if format == features.BoundingBoxFormat.XYXY:
bounding_box[..., 2] += left
bounding_box[..., 3] += top
return bounding_box


crop_image_tensor = _FT.crop
Expand Down