Skip to content

Port random affine, rotate, perspective and to_grayscale to pytest #4000

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
Jun 9, 2021
Merged

Port random affine, rotate, perspective and to_grayscale to pytest #4000

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
c4b457d
Port random affine, rotate, perspective and to_grayscale to pytest
vivekkumar7089 Jun 8, 2021
19047cb
Defined separate save function
vivekkumar7089 Jun 9, 2021
e0f60f0
Merging to update it from master
vivekkumar7089 Jun 9, 2021
aa64908
Merge branch 'master' of github.com:pytorch/vision into port-test_tra…
NicolasHug Jun 9, 2021
c1141f4
pep8
NicolasHug Jun 9, 2021
aa4a1ae
again
NicolasHug Jun 9, 2021
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
216 changes: 121 additions & 95 deletions test/test_transforms_tensor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
_create_data_batch,
_assert_equal_tensor_to_pil,
_assert_approx_equal_tensor_to_pil,
cpu_and_gpu
cpu_and_gpu,
)
from _assert_utils import assert_equal

Expand Down Expand Up @@ -286,100 +286,6 @@ def test_resized_crop(self):
with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_resized_crop.pt"))

def test_random_affine(self):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=self.device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=self.device)

def _test(**kwargs):
transform = T.RandomAffine(**kwargs)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)

return s_transform

for interpolation in [NEAREST, BILINEAR]:
for shear in [15, 10.0, (5.0, 10.0), [-15, 15], [-10.0, 10.0, -11.0, 11.0]]:
_test(degrees=0.0, interpolation=interpolation, shear=shear)

for scale in [(0.7, 1.2), [0.7, 1.2]]:
_test(degrees=0.0, interpolation=interpolation, scale=scale)

for translate in [(0.1, 0.2), [0.2, 0.1]]:
_test(degrees=0.0, interpolation=interpolation, translate=translate)

for degrees in [45, 35.0, (-45, 45), [-90.0, 90.0]]:
_test(degrees=degrees, interpolation=interpolation)

for fill in [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1]:
_test(degrees=0.0, interpolation=interpolation, fill=fill)

s_transform = _test(degrees=0.0)
with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_random_affine.pt"))

def test_random_rotate(self):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=self.device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=self.device)

for center in [(0, 0), [10, 10], None, (56, 44)]:
for expand in [True, False]:
for degrees in [45, 35.0, (-45, 45), [-90.0, 90.0]]:
for interpolation in [NEAREST, BILINEAR]:
for fill in [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1]:
transform = T.RandomRotation(
degrees=degrees, interpolation=interpolation, expand=expand, center=center, fill=fill
)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)

with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_random_rotate.pt"))

def test_random_perspective(self):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=self.device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=self.device)

for distortion_scale in np.linspace(0.1, 1.0, num=20):
for interpolation in [NEAREST, BILINEAR]:
for fill in [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1]:
transform = T.RandomPerspective(
distortion_scale=distortion_scale,
interpolation=interpolation,
fill=fill
)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)

with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_perspective.pt"))

def test_to_grayscale(self):

meth_kwargs = {"num_output_channels": 1}
tol = 1.0 + 1e-10
_test_class_op(
T.Grayscale, meth_kwargs=meth_kwargs, test_exact_match=False, device=self.device,
tol=tol, agg_method="max"
)

meth_kwargs = {"num_output_channels": 3}
_test_class_op(
T.Grayscale, meth_kwargs=meth_kwargs, test_exact_match=False, device=self.device,
tol=tol, agg_method="max"
)

meth_kwargs = {}
_test_class_op(
T.RandomGrayscale, meth_kwargs=meth_kwargs, test_exact_match=False, device=self.device,
tol=tol, agg_method="max"
)

def test_normalize(self):
fn = T.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
tensor, _ = _create_data(26, 34, device=self.device)
Expand Down Expand Up @@ -736,5 +642,125 @@ def setUp(self):
self.device = "cuda"


def _test_random_affine_helper(device, **kwargs):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=device)
transform = T.RandomAffine(**kwargs)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)


@pytest.mark.parametrize('device', cpu_and_gpu())
def test_random_affine(device):
transform = T.RandomAffine(degrees=45.0)
s_transform = torch.jit.script(transform)
with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_random_affine.pt"))


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('shear', [15, 10.0, (5.0, 10.0), [-15, 15], [-10.0, 10.0, -11.0, 11.0]])
def test_random_affine_shear(device, interpolation, shear):
_test_random_affine_helper(device, degrees=0.0, interpolation=interpolation, shear=shear)


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('scale', [(0.7, 1.2), [0.7, 1.2]])
def test_random_affine_scale(device, interpolation, scale):
_test_random_affine_helper(device, degrees=0.0, interpolation=interpolation, scale=scale)


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('translate', [(0.1, 0.2), [0.2, 0.1]])
def test_random_affine_translate(device, interpolation, translate):
_test_random_affine_helper(device, degrees=0.0, interpolation=interpolation, translate=translate)


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('degrees', [45, 35.0, (-45, 45), [-90.0, 90.0]])
def test_random_affine_degrees(device, interpolation, degrees):
_test_random_affine_helper(device, degrees=degrees, interpolation=interpolation)


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('fill', [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1])
def test_random_affine_fill(device, interpolation, fill):
_test_random_affine_helper(device, degrees=0.0, interpolation=interpolation, fill=fill)


@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('center', [(0, 0), [10, 10], None, (56, 44)])
@pytest.mark.parametrize('expand', [True, False])
@pytest.mark.parametrize('degrees', [45, 35.0, (-45, 45), [-90.0, 90.0]])
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('fill', [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1])
def test_random_rotate(device, center, expand, degrees, interpolation, fill):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=device)

transform = T.RandomRotation(
degrees=degrees, interpolation=interpolation, expand=expand, center=center, fill=fill
)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)


def test_random_rotate_save():
transform = T.RandomRotation(degrees=45.0)
s_transform = torch.jit.script(transform)
with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_random_rotate.pt"))
Comment on lines +720 to +721
Copy link
Member

Choose a reason for hiding this comment

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

Let's move this into a separate test: with all the parametrization, it's called many many times and adds up about 4 seconds to the execution, which is no negligible. We just need to call it once in a test called e.g. test_random_rotate_save

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Hi @NicolasHug !! I have changed everything, the rest of this. Actually, I couldn't find a way to reuse s_transform outside of its parent function. Could you please help me to solve this problem?

Copy link
Member

Choose a reason for hiding this comment

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

I made a comment above, hopefully this helps? LMK!

Copy link
Contributor Author

Choose a reason for hiding this comment

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

It means here I can define a function.

def test_random_rotate_save():
    s_transform = torch.jit.scriptT.RandomAffine(...)
    with get_tmp_dir() as tmp_dir:     
         s_transform.save(os.path.join(tmp_dir, "t_random_rotate.pt")

And also in a similar manner in the function test_random_perspective.
@NicolasHug

Copy link
Member

Choose a reason for hiding this comment

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

Yes this sounds like the right way here!

Copy link
Contributor Author

@vivekkumar7089 vivekkumar7089 Jun 8, 2021 •
edited
Loading

Choose a reason for hiding this comment

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

Hello @NicolasHug !!
s_transform = torch.jit.script(T.RandomAffine())
and
s_transform = torch.jit.script(T.RandomRotation())
both are giving same error -: TypeError: init() missing 1 required positional argument: 'degrees'
while
s_transform = torch.jit.script(T.RandomPerspective()) is working fine.
Can we fix some degrees here? (in above cases)

Copy link
Member

@NicolasHug NicolasHug Jun 9, 2021 •
edited
Loading

Choose a reason for hiding this comment

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

yes we can just pass degree=0
EDIT - or 45 as you did!



@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('distortion_scale', np.linspace(0.1, 1.0, num=20))
@pytest.mark.parametrize('interpolation', [NEAREST, BILINEAR])
@pytest.mark.parametrize('fill', [85, (10, -10, 10), 0.7, [0.0, 0.0, 0.0], [1, ], 1])
def test_random_perspective(device, distortion_scale, interpolation, fill):
tensor = torch.randint(0, 256, size=(3, 44, 56), dtype=torch.uint8, device=device)
batch_tensors = torch.randint(0, 256, size=(4, 3, 44, 56), dtype=torch.uint8, device=device)

transform = T.RandomPerspective(
distortion_scale=distortion_scale,
interpolation=interpolation,
fill=fill
)
s_transform = torch.jit.script(transform)

_test_transform_vs_scripted(transform, s_transform, tensor)
_test_transform_vs_scripted_on_batch(transform, s_transform, batch_tensors)


def test_random_perspective_save():
transform = T.RandomPerspective()
s_transform = torch.jit.script(transform)
with get_tmp_dir() as tmp_dir:
s_transform.save(os.path.join(tmp_dir, "t_perspective.pt"))
Comment on lines +746 to +747
Copy link
Member

Choose a reason for hiding this comment

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

let's move this out as well



@pytest.mark.parametrize('device', cpu_and_gpu())
@pytest.mark.parametrize('Klass, meth_kwargs', [
(T.Grayscale, {"num_output_channels": 1}),
(T.Grayscale, {"num_output_channels": 3}),
(T.RandomGrayscale, {})
])
def test_to_grayscale(device, Klass, meth_kwargs):

tol = 1.0 + 1e-10
_test_class_op(
Klass, meth_kwargs=meth_kwargs, test_exact_match=False, device=device,
tol=tol, agg_method="max"
)


if __name__ == '__main__':
unittest.main()