Rewrite test and fix masks_to_boxes implementation (#4469)

Co-authored-by: Nicolas Hug <[email protected]>

test/test_ops.py

@@ -4,7 +4,9 @@
 import pytest
 
 import numpy as np
+import os
 
+from PIL import Image
 import torch
 from functools import lru_cache
 from torch import Tensor
@@ -1000,6 +1002,38 @@ def gen_iou_check(box, expected, tolerance=1e-4):
             gen_iou_check(box_tensor, expected, tolerance=0.002 if dtype == torch.float16 else 1e-3)
 
 
+class TestMasksToBoxes:
+    def test_masks_box(self):
+        def masks_box_check(masks, expected, tolerance=1e-4):
+            out = ops.masks_to_boxes(masks)
+            assert out.dtype == torch.float
+            torch.testing.assert_close(out, expected, rtol=0.0, check_dtype=False, atol=tolerance)
+
+        # Check for int type boxes.
+        def _get_image():
+            assets_directory = os.path.join(os.path.dirname(os.path.abspath(__file__)), "assets")
+            mask_path = os.path.join(assets_directory, "masks.tiff")
+            image = Image.open(mask_path)
+            return image
+
+        def _create_masks(image, masks):
+            for index in range(image.n_frames):
+                image.seek(index)
+                frame = np.array(image)
+                masks[index] = torch.tensor(frame)
+
+            return masks
+
+        expected = torch.tensor([[127, 2, 165, 40], [2, 50, 44, 92], [56, 63, 98, 100], [139, 68, 175, 104],
+                                 [160, 112, 198, 145], [49, 138, 99, 182], [108, 148, 152, 213]], dtype=torch.float)
+
+        image = _get_image()
+        for dtype in [torch.float16, torch.float32, torch.float64]:
+            masks = torch.zeros((image.n_frames, image.height, image.width), dtype=dtype)
+            masks = _create_masks(image, masks)
+            masks_box_check(masks, expected)
+
+
 class TestStochasticDepth:
     @pytest.mark.parametrize('p', [0.2, 0.5, 0.8])
     @pytest.mark.parametrize('mode', ["batch", "row"])

torchvision/ops/boxes.py

@@ -301,24 +301,24 @@ def generalized_box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
 
 def masks_to_boxes(masks: torch.Tensor) -> torch.Tensor:
     """
-    Compute the bounding boxes around the provided masks
+    Compute the bounding boxes around the provided masks.
 
-    Returns a [N, 4] tensor.  Both sets of boxes are expected to be in ``(x1, y1, x2, y2)`` format with
+    Returns a [N, 4] tensor containing bounding boxes. The boxes are in ``(x1, y1, x2, y2)`` format with
     ``0 <= x1 < x2`` and ``0 <= y1 < y2``.
 
     Args:
-        masks (Tensor[N, H, W]): masks to transform where N is the number of
-        masks and (H, W) are the spatial dimensions.
+        masks (Tensor[N, H, W]): masks to transform where N is the number of masks
+            and (H, W) are the spatial dimensions.
 
     Returns:
         Tensor[N, 4]: bounding boxes
     """
     if masks.numel() == 0:
-        return torch.zeros((0, 4))
+        return torch.zeros((0, 4), device=masks.device, dtype=torch.float)
 
     n = masks.shape[0]
 
-    bounding_boxes = torch.zeros((n, 4), device=masks.device, dtype=torch.int)
+    bounding_boxes = torch.zeros((n, 4), device=masks.device, dtype=torch.float)
 
     for index, mask in enumerate(masks):
         y, x = torch.where(masks[index] != 0)