update doc

README.md

 # tensorpack
 Neural Network Toolbox on TensorFlow
 
-Still in development, but usable.
-
 See some [examples](examples) to learn about the framework.
-You can actually train them and reproduce the performance... not just see how to write code.
+You can actually train them and reproduce the performance... not just to see how to write code.
 
 + [DoReFa-Net: training binary / low bitwidth CNN](examples/DoReFa-Net)
 + [IncpetionV3 on ImageNet](examples/Inception/inceptionv3.py)
@@ -17,16 +15,16 @@ You can actually train them and reproduce the performance... not just see how to
 
 ## Features:
 
-Abstract your training task into three components:
+Describe your training task with three components:
 
-1. Model, or graph. `models/` has some scoped abstraction of common models. This part is roughly an
-	 equivalent of slim/tflearn/tensorlayer.
+1. Model, or graph. `models/` has some scoped abstraction of common models, but you can simply use
+	 anything available in tensorflow. This part is roughly an equivalent of slim/tflearn/tensorlayer.
 	`LinearWrap` and `argscope` makes large models look simpler.
 
 2. Data. tensorpack allows and encourages complex data processing.
 
-	+ All data producer has an unified `DataFlow` interface, allowing them to be composed to perform complex preprocessing.
-	+ Use Python to easily handle your own data format, yet still keep a good training speed thanks to multiprocess prefetch & TF Queue prefetch.
+	+ All data producer has an unified `generator` interface, allowing them to be composed to perform complex preprocessing.
+	+ Use Python to easily handle any of your own data format, yet still keep a good training speed thanks to multiprocess prefetch & TF Queue prefetch.
 	For example, InceptionV3 can run in the same speed as the official code which reads data using TF operators.
 
 3. Callbacks, including everything you want to do apart from the training iterations. Such as:

examples/SpatialTransformer/README.md

-## Spatial Transformer Network
+## Spatial Transformer Networks
 
 Reproduce the "MNIST addition" experiments in
 [Spatial Transformer Networks](https://arxiv.org/abs/1506.02025)
@@ -10,7 +10,8 @@ Here the two Spatial Transformer branches learn to localize the two digits
 and warped them separately.
 
 ![demo](demo.jpg)
-(Left: input image; Middle: output of the first STN branch (which localizes the second digit); Right: output of the second STN branch.)
+
+Left: input image; Middle: output of the first STN branch (which localizes the second digit); Right: output of the second STN branch.
 
 To train (takes about 300 epochs to reach 8.8% error):
 ```bash