Merge pull request #1503 from damienv-gh/master

Entropy coder for images: remove deprecated functions and update README.

compression/README.md

@@ -8,6 +8,7 @@ code for the following papers:
 
 ## Organization
 [Image Encoder](image_encoder/): Encoding and decoding images into their binary representation.
+[Entropy Coder](entropy_coder/): Lossless compression of the binary representation.
 
 ## Contact Info
 Model repository maintained by Nick Johnston ([nickj-google](https://github.com/nickj-google)).

compression/entropy_coder/README.md

@@ -14,6 +14,11 @@ the width of the binary codes,
 sliced into N groups of K, where each additional group is used by the image
 decoder to add more details to the reconstructed image.
 
+The code in this directory only contains the underlying code probability model
+but does not perform the actual compression using arithmetic coding.
+The code probability model is enough to compute the theoretical compression
+ratio.
+
 
 ## Prerequisites
 The only software requirements for running the encoder and decoder is having
@@ -22,7 +27,7 @@ Tensorflow installed.
 You will also need to add the top level source directory of the entropy coder
 to your `PYTHONPATH`, for example:
 
-`export PYTHONPATH=${PYTHONPATH}:/tmp/compression/entropy_coder`
+`export PYTHONPATH=${PYTHONPATH}:/tmp/models/compression`
 
 
 ## Training the entropy coder
@@ -38,6 +43,8 @@ less.
 
 To generate a synthetic dataset with 20000 samples:
 
+`mkdir -p /tmp/dataset`
+
 `python ./dataset/gen_synthetic_dataset.py --dataset_dir=/tmp/dataset/
 --count=20000`
 

compression/entropy_coder/core/entropy_coder_train.py

@@ -111,7 +111,7 @@ def train():
           decay_steps=decay_steps,
           decay_rate=decay_rate,
           staircase=True)
-      tf.contrib.deprecated.scalar_summary('Learning Rate', learning_rate)
+      tf.summary.scalar('Learning Rate', learning_rate)
       optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate,
                                          epsilon=1.0)
 

compression/entropy_coder/progressive/progressive.py

@@ -202,11 +202,10 @@ class ProgressiveModel(entropy_coder_model.EntropyCoderModel):
       code_length.append(code_length_block(
           blocks.ConvertSignCodeToZeroOneCode(x),
           blocks.ConvertSignCodeToZeroOneCode(predicted_x)))
-      tf.contrib.deprecated.scalar_summary('code_length_layer_{:02d}'.format(k),
-                                           code_length[-1])
+      tf.summary.scalar('code_length_layer_{:02d}'.format(k), code_length[-1])
     code_length = tf.stack(code_length)
     self.loss = tf.reduce_mean(code_length)
-    tf.contrib.deprecated.scalar_summary('loss', self.loss)
+    tf.summary.scalar('loss', self.loss)
 
     # Loop over all the remaining layers just to make sure they are
     # instantiated. Otherwise, loading model params could fail.