diff --git a/apps/trainers/FaceAligner/TrainingDataGenerator.py b/apps/trainers/FaceAligner/TrainingDataGenerator.py
index e6f20b3..6d93d19 100644
--- a/apps/trainers/FaceAligner/TrainingDataGenerator.py
+++ b/apps/trainers/FaceAligner/TrainingDataGenerator.py
@@ -196,7 +196,7 @@ class TrainingDataGenerator(lib_mp.MPWorker):
                         ip = lib_img.ImageProcessor(img_aligned_shifted)
                         rnd = np.random
                         if rnd.randint(2) == 0:
-                            ip.hsv( rnd.randint(0, 360), rnd.uniform(-0.5,0.5), rnd.uniform(-0.5,0.5), mask=lib_sd.random_circle_faded_multi((resolution,resolution), complexity=random_mask_complexity))
+                            ip.hsv( rnd.uniform(0,1), rnd.uniform(-0.5,0.5), rnd.uniform(-0.5,0.5), mask=lib_sd.random_circle_faded_multi((resolution,resolution), complexity=random_mask_complexity))
                         else:
                             ip.levels( [ [rnd.uniform(0,0.25),rnd.uniform(0.75,1.0),rnd.uniform(0.5,1.5), rnd.uniform(0,0.25),rnd.uniform(0.75,1.0),  ],
                                          [rnd.uniform(0,0.25),rnd.uniform(0.75,1.0),rnd.uniform(0.5,1.5), rnd.uniform(0,0.25),rnd.uniform(0.75,1.0),],
diff --git a/xlib/image/ImageProcessor.py b/xlib/image/ImageProcessor.py
index 2ffb6b5..8e8a1ea 100644
--- a/xlib/image/ImageProcessor.py
+++ b/xlib/image/ImageProcessor.py
@@ -466,9 +466,9 @@ class ImageProcessor:
         """
         apply HSV modification for BGR image
             
-            h_diff = [-360.0 .. 360.0]
-            s_diff = [-1.0 .. 1.0]
+            h_diff = [-1.0 .. 1.0]
             s_diff = [-1.0 .. 1.0]
+            v_diff = [-1.0 .. 1.0]
         """
         dtype = self.get_dtype()
         self.to_ufloat32()
@@ -481,7 +481,7 @@ class ImageProcessor:
         img = img.reshape( (N*H,W,C) )
         
         h, s, v = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))
-        h = ( h + h_diff ) % 360
+        h = ( h + h_diff*360.0 ) % 360
         
         s += s_diff
         np.clip (s, 0, 1, out=s )