summary / dataflow fix

tensorpack/dataflow/common.py

@@ -130,7 +130,7 @@ class FakeData(DataFlow):
 class MapData(ProxyDataFlow):
     """ Map a function to the datapoint"""
     def __init__(self, ds, func):
-        super(MapData, self).__init_(ds)
+        super(MapData, self).__init__(ds)
         self.func = func
 
     def get_data(self):

tensorpack/dataflow/image.py

@@ -7,6 +7,7 @@ import numpy as np
 import cv2
 import copy
 from .base import DataFlow, ProxyDataFlow
+from .common import MapDataComponent
 from .imgaug import AugmentorList, Image
 
 __all__ = ['ImageFromFile', 'AugmentImageComponent']
@@ -37,7 +38,7 @@ class ImageFromFile(DataFlow):
             yield [im]
 
 
-class AugmentImageComponent(ProxyDataFlow):
+class AugmentImageComponent(MapDataComponent):
     """
     Augment image in each data point
     Args:
@@ -46,16 +47,10 @@ class AugmentImageComponent(ProxyDataFlow):
         index: the index of image in each data point. default to be 0
     """
     def __init__(self, ds, augmentors, index=0):
-        super(AugmentImageComponent, self).__init__(ds)
         self.augs = AugmentorList(augmentors)
-        self.index = index
+        super(AugmentImageComponent, self).__init__(
+            ds, lambda x: self.augs.augment(Image(x)).arr, index)
 
     def reset_state(self):
         self.ds.reset_state()
         self.augs.reset_state()
-
-    def get_data(self):
-        for dp in self.ds.get_data():
-            dp = copy.deepcopy(dp)
-            dp[self.index] = self.augs.augment(Image(dp[self.index])).arr
-            yield dp

tensorpack/utils/summary.py

@@ -5,7 +5,7 @@
 
 import tensorflow as tf
 
-from . import logger
+from . import logger, get_global_step_var
 from .naming import *
 
 def create_summary(name, v):
@@ -30,8 +30,8 @@ def add_activation_summary(x, name=None):
         "Summary a scalar with histogram? Maybe use scalar instead. FIXME!"
     if name is None:
         name = x.name
-    tf.histogram_summary(name + '/activations', x)
-    tf.scalar_summary(name + '/sparsity', tf.nn.zero_fraction(x))
+    tf.histogram_summary(name + '/activation', x)
+    tf.scalar_summary(name + '/activation_sparsity', tf.nn.zero_fraction(x))
 
 def add_param_summary(regex):
     """
@@ -45,6 +45,7 @@ def add_param_summary(regex):
             if p.get_shape().ndims == 0:
                 tf.scalar_summary(name, p)
             else:
+                tf.scalar_summary(name + '/sparsity', tf.nn.zero_fraction(p))
                 tf.histogram_summary(name, p)
 
 def summary_moving_average(cost_var):
@@ -52,7 +53,7 @@ def summary_moving_average(cost_var):
         MOVING_SUMMARY_VARS_KEY, as well as the argument
         Return a op to maintain these average
     """
-    global_step_var = tf.get_default_graph().get_tensor_by_name(GLOBAL_STEP_VAR_NAME)
+    global_step_var = get_global_step_var()
     averager = tf.train.ExponentialMovingAverage(
         0.99, num_updates=global_step_var, name='moving_averages')
     vars_to_summary = [cost_var] + \