centerpatch dataflow

tensorpack/dataflow/imgaug/crop.py

@@ -4,7 +4,10 @@
 
 from .base import ImageAugmentor
 
-__all__ = ['RandomCrop', 'CenterCrop', 'FixedCrop']
+import numpy as np
+from abc import abstractmethod
+
+__all__ = ['RandomCrop', 'CenterCrop', 'FixedCrop', 'CenterPaste']
 
 class RandomCrop(ImageAugmentor):
     """ Randomly crop the image into a smaller one """
@@ -47,3 +50,47 @@ class FixedCrop(ImageAugmentor):
                           self.rangex[0]:self.rangex[1]]
         if img.coords:
             raise NotImplementedError()
+
+class BackgroundFiller(object):
+    @abstractmethod
+    def fill(background_shape, img):
+        """
+        return a proper background image of background_shape, given img
+        """
+
+class ConstantBackgroundFiller(BackgroundFiller):
+    def __init__(self, value):
+        self.value = value
+
+    def fill(self, background_shape, img):
+        assert img.ndim in [3, 1]
+        if img.ndim == 3:
+            return_shape = background_shape + (3,)
+        else:
+            return_shape = background_shape
+        return np.zeros(return_shape) + self.value
+
+class CenterPaste(ImageAugmentor):
+    """
+    Paste the image onto center of a background
+    """
+    def __init__(self, background_shape, background_filler=None):
+        if background_filler is None:
+            background_filler = ConstantBackgroundFiller(0)
+
+        self._init(locals())
+
+    def _augment(self, img):
+        img_shape = img.arr.shape[:2]
+        assert self.background_shape[0] > img_shape[0] and self.background_shape[1] > img_shape[1]
+
+        background = self.background_filler.fill(
+            self.background_shape, img.arr)
+        h0 = (self.background_shape[0] - img_shape[0]) * 0.5
+        w0 = (self.background_shape[1] - img_shape[1]) * 0.5
+        background[h0:h0+img_shape[0], w0:w0+img_shape[1]] = img.arr
+        img.arr = background
+        if img.coords:
+            raise NotImplementedError()
+
+

tensorpack/dataflow/imgaug/imgproc.py

@@ -11,13 +11,14 @@ class BrightnessAdd(ImageAugmentor):
     """
     Randomly add a value within [-delta,delta], and clip in [0,255]
     """
-    def __init__(self, delta):
+    def __init__(self, delta, clip=True):
         assert delta > 0
         self._init(locals())
 
     def _augment(self, img):
         v = self._rand_range(-self.delta, self.delta)
         img.arr += v
+        if self.clip:
             img.arr = np.clip(img.arr, 0, 255)
 
 class Contrast(ImageAugmentor):
@@ -25,7 +26,7 @@ class Contrast(ImageAugmentor):
     Apply x = (x - mean) * contrast_factor + mean to each channel
     and clip to [0, 255]
     """
-    def __init__(self, factor_range):
+    def __init__(self, factor_range, clip=True):
         self._init(locals())
 
     def _augment(self, img):
@@ -33,6 +34,7 @@ class Contrast(ImageAugmentor):
         r = self._rand_range(*self.factor_range)
         mean = np.mean(arr, axis=(0,1), keepdims=True)
         img.arr = (arr - mean) * r + mean
+        if self.clip:
             img.arr = np.clip(img.arr, 0, 255)
 
 class MeanVarianceNormalize(ImageAugmentor):

tensorpack/models/nonlin.py

 #!/usr/bin/env python2
 # -*- coding: UTF-8 -*-
-# File: nonl.py
+# File: nonlin.py
 # Author: Yuxin Wu <ppwwyyxx@gmail.com>
 
 import tensorflow as tf
@@ -8,7 +8,7 @@ from copy import copy
 
 from ._common import *
 
-__all__ = ['Maxout', 'PReLU']
+__all__ = ['Maxout', 'PReLU', 'LeakyReLU']
 
 @layer_register()
 def Maxout(x, num_unit):
@@ -28,3 +28,12 @@ def PReLU(x, init=tf.constant_initializer(0.001), name=None):
         return x * 0.5
     else:
         return tf.mul(x, 0.5, name=name)
+
+@layer_register()
+def LeakyReLU(x, alpha, name=None):
+    alpha = float(alpha)
+    x = ((1 + alpha) * x + (1 - alpha) * tf.abs(x))
+    if name is None:
+        return x * 0.5
+    else:
+        return tf.mul(x, 0.5, name=name)