Add rgb=None in "imgaug.Contrast"

docs/tutorial/trainer.md

@@ -62,7 +62,7 @@ follow these conventions and will need some workarounds if used within tensorpac
 
 #### What You Can Do Inside Tower Function
 1. Call any symbolic functions as long as they follow the above rules.
-2. The function will be called under a
+2. The tower function will be called under a
  [TowerContext](../modules/tfutils.html#tensorpack.tfutils.tower.BaseTowerContext),
  which can be accessed by [get_current_tower_context()](../modules/tfutils.html#tensorpack.tfutils.tower.get_current_tower_context).
    The context contains information about training/inference mode, scope name, etc.
@@ -93,7 +93,7 @@ Note some __common problems__ when using these trainers:
         inputs on each GPU needs to have consistent shapes.
     ```
 
-2. The tower function (your model code) will get called multipile times on each GPU.
+2. The tower function (your model code) will get called once on each GPU.
    You must follow the abovementieond rules of tower function.
 
 ### Distributed Trainers

tensorpack/dataflow/imgaug/imgproc.py

@@ -51,7 +51,7 @@ class Brightness(ImageAugmentor):
         """
         Args:
             delta (float): Randomly add a value within [-delta,delta]
-            clip (bool): clip results to [0,255].
+            clip (bool): clip results to [0,255] if data type is uint8.
         """
         super(Brightness, self).__init__()
         assert delta > 0
@@ -78,7 +78,7 @@ class BrightnessScale(ImageAugmentor):
         """
         Args:
             range (tuple): Randomly scale the image by a factor in (range[0], range[1])
-            clip (bool): clip results to [0,255].
+            clip (bool): clip results to [0,255] if data type is uint8.
         """
         super(BrightnessScale, self).__init__()
         self._init(locals())
@@ -101,11 +101,12 @@ class Contrast(ImageAugmentor):
     Apply ``x = (x - mean) * contrast_factor + mean`` to each channel.
     """
 
-    def __init__(self, factor_range, rgb=True, clip=True):
+    def __init__(self, factor_range, rgb=None, clip=True):
         """
         Args:
             factor_range (list or tuple): an interval to randomly sample the `contrast_factor`.
-            clip (bool): clip to [0, 255] if True.
+            rgb (bool or None): if None, use the mean per-channel.
+            clip (bool): clip to [0, 255] if data type is uint8.
         """
         super(Contrast, self).__init__()
         self._init(locals())
@@ -117,9 +118,12 @@ class Contrast(ImageAugmentor):
         old_dtype = img.dtype
 
         if img.ndim == 3:
-            m = cv2.COLOR_RGB2GRAY if self.rgb else cv2.COLOR_BGR2GRAY
-            grey = cv2.cvtColor(img, m)
-            mean = np.mean(grey)
+            if self.rgb is not None:
+                m = cv2.COLOR_RGB2GRAY if self.rgb else cv2.COLOR_BGR2GRAY
+                grey = cv2.cvtColor(img.astype('float32'), m)
+                mean = np.mean(grey)
+            else:
+                mean = np.mean(img, axis=(0, 1), keepdims=True)
         else:
             mean = np.mean(img)