fix transforms/bbox

chainercv/transforms/bbox/flip_bbox.py

@@ -1,10 +1,10 @@
-def flip_bbox(bbox, size, x_flip=False, y_flip=False):
+def flip_bbox(bbox, size, y_flip=False, x_flip=False):
     """Flip bounding boxes accordingly.
 
     The bounding boxes are expected to be packed into a two dimensional
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
-    the bounding box. They are :obj:`(x_min, y_min, x_max, y_max)`,
+    the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
     where the four attributes are coordinates of the bottom left and the
     top right vertices.
 
@@ -13,10 +13,10 @@ def flip_bbox(bbox, size, x_flip=False, y_flip=False):
             :math:`R` is the number of bounding boxes.
         size (tuple): A tuple of length 2. The height and the width
             of the image before resized.
-        x_flip (bool): Flip bounding box according to a horizontal flip of
-            an image.
         y_flip (bool): Flip bounding box according to a vertical flip of
             an image.
+        x_flip (bool): Flip bounding box according to a horizontal flip of
+            an image.
 
     Returns:
         ~numpy.ndarray:
@@ -25,14 +25,14 @@ def flip_bbox(bbox, size, x_flip=False, y_flip=False):
     """
     H, W = size
     bbox = bbox.copy()
-    if x_flip:
-        x_max = W - 1 - bbox[:, 0]
-        x_min = W - 1 - bbox[:, 2]
-        bbox[:, 0] = x_min
-        bbox[:, 2] = x_max
     if y_flip:
-        y_max = H - 1 - bbox[:, 1]
-        y_min = H - 1 - bbox[:, 3]
-        bbox[:, 1] = y_min
-        bbox[:, 3] = y_max
+        y_max = H - 1 - bbox[:, 0]
+        y_min = H - 1 - bbox[:, 2]
+        bbox[:, 0] = y_min
+        bbox[:, 2] = y_max
+    if x_flip:
+        x_max = W - 1 - bbox[:, 1]
+        x_min = W - 1 - bbox[:, 3]
+        bbox[:, 1] = x_min
+        bbox[:, 3] = x_max
     return bbox

chainercv/transforms/bbox/resize_bbox.py

@@ -4,7 +4,7 @@ def resize_bbox(bbox, in_size, out_size):
     The bounding boxes are expected to be packed into a two dimensional
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
-    the bounding box. They are :obj:`(x_min, y_min, x_max, y_max)`,
+    the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
     where the four attributes are coordinates of the bottom left and the
     top right vertices.
 
@@ -22,10 +22,10 @@ def resize_bbox(bbox, in_size, out_size):
 
     """
     bbox = bbox.copy()
-    x_scale = float(out_size[1]) / in_size[1]
     y_scale = float(out_size[0]) / in_size[0]
-    bbox[:, 0] = x_scale * bbox[:, 0]
-    bbox[:, 2] = x_scale * bbox[:, 2]
-    bbox[:, 1] = y_scale * bbox[:, 1]
-    bbox[:, 3] = y_scale * bbox[:, 3]
+    x_scale = float(out_size[1]) / in_size[1]
+    bbox[:, 0] = y_scale * bbox[:, 0]
+    bbox[:, 2] = y_scale * bbox[:, 2]
+    bbox[:, 1] = x_scale * bbox[:, 1]
+    bbox[:, 3] = x_scale * bbox[:, 3]
     return bbox

chainercv/transforms/bbox/translate_bbox.py

@@ -1,22 +1,23 @@
-def translate_bbox(bbox, x_offset=0, y_offset=0):
+def translate_bbox(bbox, y_offset=0, x_offset=0):
     """Translate bounding boxes.
 
     This method is mainly used together with image transforms, such as padding
     and cropping, which translates the left top point of the image from
-    coordinate :math:`(0, 0)` to coordinate :math:`(x\_offset, y\_offset)`.
+    coordinate :math:`(0, 0)` to coordinate
+    :math:`(y, x) = (y\_offset, x\_offset)`.
 
     The bounding boxes are expected to be packed into a two dimensional
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
-    the bounding box. They are :obj:`(x_min, y_min, x_max, y_max)`,
+    the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
     where the four attributes are coordinates of the bottom left and the
     top right vertices.
 
     Args:
         bbox (~numpy.ndarray): Bounding boxes to be transformed. The shape is
             :math:`(R, 4)`. :math:`R` is the number of bounding boxes.
-        x_offset (int or float): The offset along x axis.
         y_offset (int or float): The offset along y axis.
+        x_offset (int or float): The offset along x axis.
 
     Returns:
         ~numpy.ndarray:
@@ -25,7 +26,7 @@ def translate_bbox(bbox, x_offset=0, y_offset=0):
     """
 
     out_bbox = bbox.copy()
-    out_bbox[:, :2] += (x_offset, y_offset)
-    out_bbox[:, 2:] += (x_offset, y_offset)
+    out_bbox[:, :2] += (y_offset, x_offset)
+    out_bbox[:, 2:] += (y_offset, x_offset)
 
     return out_bbox

tests/transforms_tests/bbox_tests/test_flip_bbox.py

@@ -12,16 +12,16 @@ def test_flip_bbox(self):
         bbox = np.random.uniform(
             low=0., high=32., size=(10, 4))
 
-        out = flip_bbox(bbox, size=(34, 32), x_flip=True)
+        out = flip_bbox(bbox, size=(34, 32), y_flip=True)
         bbox_expected = bbox.copy()
-        bbox_expected[:, 0] = 31 - bbox[:, 2]
-        bbox_expected[:, 2] = 31 - bbox[:, 0]
+        bbox_expected[:, 0] = 33 - bbox[:, 2]
+        bbox_expected[:, 2] = 33 - bbox[:, 0]
         np.testing.assert_equal(out, bbox_expected)
 
-        out = flip_bbox(bbox, size=(34, 32), y_flip=True)
+        out = flip_bbox(bbox, size=(34, 32), x_flip=True)
         bbox_expected = bbox.copy()
-        bbox_expected[:, 1] = 33 - bbox[:, 3]
-        bbox_expected[:, 3] = 33 - bbox[:, 1]
+        bbox_expected[:, 1] = 31 - bbox[:, 3]
+        bbox_expected[:, 3] = 31 - bbox[:, 1]
         np.testing.assert_equal(out, bbox_expected)
 
 

tests/transforms_tests/bbox_tests/test_resize_bbox.py

@@ -14,10 +14,10 @@ def test_resize_bbox(self):
 
         out = resize_bbox(bbox, in_size=(32, 32), out_size=(64, 128))
         bbox_expected = bbox.copy()
-        bbox_expected[:, 0] = bbox[:, 0] * 4
-        bbox_expected[:, 1] = bbox[:, 1] * 2
-        bbox_expected[:, 2] = bbox[:, 2] * 4
-        bbox_expected[:, 3] = bbox[:, 3] * 2
+        bbox_expected[:, 0] = bbox[:, 0] * 2
+        bbox_expected[:, 1] = bbox[:, 1] * 4
+        bbox_expected[:, 2] = bbox[:, 2] * 2
+        bbox_expected[:, 3] = bbox[:, 3] * 4
         np.testing.assert_equal(out, bbox_expected)
 
 

tests/transforms_tests/bbox_tests/test_translate_bbox.py

@@ -12,12 +12,12 @@ def test_translate_bbox(self):
         bbox = np.random.uniform(
             low=0., high=32., size=(10, 4))
 
-        out = translate_bbox(bbox, x_offset=3, y_offset=5)
+        out = translate_bbox(bbox, y_offset=5, x_offset=3)
         bbox_expected = np.empty_like(bbox)
-        bbox_expected[:, 0] = bbox[:, 0] + 3
-        bbox_expected[:, 1] = bbox[:, 1] + 5
-        bbox_expected[:, 2] = bbox[:, 2] + 3
-        bbox_expected[:, 3] = bbox[:, 3] + 5
+        bbox_expected[:, 0] = bbox[:, 0] + 5
+        bbox_expected[:, 1] = bbox[:, 1] + 3
+        bbox_expected[:, 2] = bbox[:, 2] + 5
+        bbox_expected[:, 3] = bbox[:, 3] + 3
         np.testing.assert_equal(out, bbox_expected)