added tensorflow graph to README.md

README.md

@@ -17,6 +17,10 @@ Used pre-trained VGG16 model from [https://github.com/machrisaa/tensorflow-vgg](
 * `train.py`: train the residual encoder model using tensorflow build-in GradientDescentOptimizer
 * `test.py`: test your own image and save the output image
 
+## Tensorflow graph
+
+![](images/residuall_encoder.png)
+
 ## How to use
 
 * First please download pre-trained VGG16 model [vgg16.npy](https://mega.nz/#!YU1FWJrA!O1ywiCS2IiOlUCtCpI6HTJOMrneN-Qdv3ywQP5poecM) to vgg folder

common.py

@@ -55,7 +55,7 @@ def init_model(train=True):
     residual_encoder = ResidualEncoder()
 
     # Color image
-    color_image_rgb = input_pipeline(file_paths, batch_size)
+    color_image_rgb = input_pipeline(file_paths, batch_size, test=not train)
     color_image_yuv = rgb_to_yuv(color_image_rgb, "rgb2yuv_for_color_image")
 
     # Gray image
@@ -85,10 +85,10 @@ def init_model(train=True):
                    exclusive=True, name="cost")
 
     # Using different learning rate in different training steps
-    lr = tf.div(learning_rate, tf.cast(tf.pow(2, tf.div(global_step, 160000)), tf.float32), name="learning_rate")
+    # lr = tf.div(learning_rate, tf.cast(tf.pow(2, tf.div(global_step, 160000)), tf.float32), name="learning_rate")
 
     # Optimizer
-    optimizer = tf.train.GradientDescentOptimizer(learning_rate=lr).minimize(cost, global_step=global_step)
+    optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate).minimize(cost, global_step=global_step)
 
     # Summaries
     print "Init summaries"

config.py

@@ -13,7 +13,7 @@
 image_size = 224
 
 # Parameters for neural network
-learning_rate = 1e-4  # Initial learning rate, every 5000 step we divide this by 2
+learning_rate = 1e-4  # Learning rate
 training_iters = 2400000  # The training iterations number
 batch_size = 1  # The batch size
 display_step = 1  # Display loss for each step