[fbsync] feat: add functional pad on segmentation mask (#5866)

Summary:
* feat: add functional pad on segmentation mask

* test: add basic correctness test with random masks

* test: add all padding options

* fix: pr comments

* fix: tests

* refactor: reshape tensor in 4d, then pad

Reviewed By: jdsgomes, NicolasHug

Differential Revision: D36095691

fbshipit-source-id: 1e31988216fea1664c1fd48ee39598d28bac8308

Co-authored-by: Federico Pozzi <[email protected]>

Author: 2 people

test/test_prototype_transforms_functional.py

@@ -370,6 +370,16 @@ def resized_crop_segmentation_mask():
         yield SampleInput(mask, top=top, left=left, height=height, width=width, size=size)
 
 
+@register_kernel_info_from_sample_inputs_fn
+def pad_segmentation_mask():
+    for mask, padding, padding_mode in itertools.product(
+        make_segmentation_masks(),
+        [[1], [1, 1], [1, 1, 2, 2]],  # padding
+        ["constant", "symmetric", "edge", "reflect"],  # padding mode,
+    ):
+        yield SampleInput(mask, padding=padding, padding_mode=padding_mode)
+
+
 @pytest.mark.parametrize(
     "kernel",
     [
@@ -1031,3 +1041,47 @@ def _compute_expected(mask, top_, left_, height_, width_, size_):
     expected_mask = _compute_expected(in_mask, top, left, height, width, size)
     output_mask = F.resized_crop_segmentation_mask(in_mask, top, left, height, width, size)
     torch.testing.assert_close(output_mask, expected_mask)
+
+
+@pytest.mark.parametrize("device", cpu_and_gpu())
+def test_correctness_pad_segmentation_mask_on_fixed_input(device):
+    mask = torch.ones((1, 3, 3), dtype=torch.long, device=device)
+
+    out_mask = F.pad_segmentation_mask(mask, padding=[1, 1, 1, 1])
+
+    expected_mask = torch.zeros((1, 5, 5), dtype=torch.long, device=device)
+    expected_mask[:, 1:-1, 1:-1] = 1
+    torch.testing.assert_close(out_mask, expected_mask)
+
+
+@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
+
+        h, w = mask.shape[-2], mask.shape[-1]
+        pad_left, pad_up, pad_right, pad_down = parse_padding()
+
+        new_h = h + pad_up + pad_down
+        new_w = w + pad_left + pad_right
+
+        new_shape = (*mask.shape[:-2], new_h, new_w) if len(mask.shape) > 2 else (new_h, new_w)
+        expected_mask = torch.zeros(new_shape, dtype=torch.long)
+        expected_mask[..., pad_up:-pad_down, pad_left:-pad_right] = mask
+
+        return expected_mask
+
+    for mask in make_segmentation_masks():
+        out_mask = F.pad_segmentation_mask(mask, padding, "constant")
+
+        expected_mask = _compute_expected_mask()
+        torch.testing.assert_close(out_mask, expected_mask)

torchvision/prototype/transforms/functional/__init__.py

@@ -62,6 +62,7 @@
     pad_bounding_box,
     pad_image_tensor,
     pad_image_pil,
+    pad_segmentation_mask,
     crop_bounding_box,
     crop_image_tensor,
     crop_image_pil,

torchvision/prototype/transforms/functional/_geometry.py

@@ -396,6 +396,20 @@ def rotate_segmentation_mask(
 pad_image_pil = _FP.pad
 
 
+def pad_segmentation_mask(
+    segmentation_mask: torch.Tensor, padding: List[int], padding_mode: str = "constant"
+) -> torch.Tensor:
+    num_masks, height, width = segmentation_mask.shape[-3:]
+    extra_dims = segmentation_mask.shape[:-3]
+
+    padded_mask = pad_image_tensor(
+        img=segmentation_mask.view(-1, num_masks, height, width), padding=padding, fill=0, padding_mode=padding_mode
+    )
+
+    new_height, new_width = padded_mask.shape[-2:]
+    return padded_mask.view(extra_dims + (num_masks, new_height, new_width))
+
+
 def pad_bounding_box(
     bounding_box: torch.Tensor, padding: List[int], format: features.BoundingBoxFormat
 ) -> torch.Tensor: