RandomResize accept range as pixels (fix #343)

tensorpack/dataflow/imgaug/noname.py

@@ -125,14 +125,16 @@ class ResizeShortestEdge(ImageAugmentor):
 
 
 class RandomResize(ImageAugmentor):
-    """ Randomly rescale w and h of the image"""
+    """ Randomly rescale width and height of the image."""
 
     def __init__(self, xrange, yrange, minimum=(0, 0), aspect_ratio_thres=0.15,
                  interp=cv2.INTER_LINEAR):
         """
         Args:
-            xrange (tuple): (min, max) range of scaling ratio for w, e.g. (0.9, 1.2)
-            yrange (tuple): (min, max) range of scaling ratio for h
+            xrange (tuple): a (min, max) tuple. If is floating point, the
+                tuple defines the range of scaling ratio of new width, e.g. (0.9, 1.2).
+                If is integer, the tuple defines the range of new width in pixels, e.g. (200, 350).
+            yrange (tuple): similar to xrange, but for height.
             minimum (tuple): (xmin, ymin) in pixels. To avoid scaling down too much.
             aspect_ratio_thres (float): discard samples which change aspect ratio
                 larger than this threshold. Set to 0 to keep aspect ratio.
@@ -144,10 +146,17 @@ class RandomResize(ImageAugmentor):
             assert xrange == yrange
         self._init(locals())
 
+        def is_float(tp):
+            return isinstance(tp[0], float) or isinstance(tp[1], float)
+        assert is_float(xrange) == is_float(yrange), "xrange and yrange has different type!"
+        self._is_scale = is_float(xrange)
+
     def _get_augment_params(self, img):
         cnt = 0
         h, w = img.shape[:2]
-        while True:
+
+        def get_dest_size():
+            if self._is_scale:
                 sx = self._rand_range(*self.xrange)
                 if self.aspect_ratio_thres == 0:
                     sy = sx
@@ -155,15 +164,29 @@ class RandomResize(ImageAugmentor):
                     sy = self._rand_range(*self.yrange)
                 destX = max(sx * w, self.minimum[0])
                 destY = max(sy * h, self.minimum[1])
+            else:
+                sx = int(self._rand_range(*self.xrange))
+                if self.aspect_ratio_thres == 0:
+                    sy = sx * 1.0 / w * h
+                else:
+                    sy = self._rand_range(*self.yrange)
+                destX = max(sx, self.minimum[0])
+                destY = max(sy, self.minimum[1])
+            return (destX, destY)
+
+        while True:
+            destX, destY = get_dest_size()
+            if self.aspect_ratio_thres > 0:  # don't check when thres == 0
                 oldr = w * 1.0 / h
                 newr = destX * 1.0 / destY
                 diff = abs(newr - oldr) / oldr
-            if diff <= self.aspect_ratio_thres + 1e-5:
-                return (h, w, int(destY), int(destX))
+                if diff >= self.aspect_ratio_thres + 1e-5:
                     cnt += 1
                     if cnt > 50:
                         logger.warn("RandomResize failed to augment an image")
                         return (h, w, h, w)
+                    continue
+            return (h, w, int(destY), int(destX))
 
     def _augment(self, img, param):
         _, _, newh, neww = param