Refactor: remove all periods in python comment

Author: Armour

README.md

@@ -15,7 +15,7 @@ This is a Tensorflow implementation of the Residual Encoder Network based on [Au
 * `read_input.py`: all functions related to input
 * `residual_encoder.py`: the residual encoder model
 * `common.py`: the common part for training and testing, which is mainly the workflow for this model
-* `train.py`: train the residual encoder model using Tensorflow built-in GradientDescentOptimizer
+* `train.py`: train the residual encoder model using Tensorflow built-in AdamOptimizer
 * `test.py`: test your own images and save the output images
 
 ## Tensorflow graph
@@ -26,11 +26,11 @@ This is a Tensorflow implementation of the Residual Encoder Network based on [Au
 
 * First please download pre-trained VGG16 model [vgg16.npy](https://mega.nz/#!YU1FWJrA!O1ywiCS2IiOlUCtCpI6HTJOMrneN-Qdv3ywQP5poecM) to vgg folder
 
-* Use pre-trained residual encoder model
+* Option 1: Use pre-trained residual encoder model
   * Model can be downloaded [here](https://github.com/Armour/Automatic-Image-Colorization/releases/tag/2.0)
   * Unzip all files to `summary_path` (you can change this path in `config.py`)
 
-* Train your own model
+* Option 2: Train your own model!
   1. Change the `batch_size` and `training_iters` if you want.
   2. Change `training_dir` to your directory that contains all your training jpg images
   3. Run `python train.py`
@@ -54,7 +54,7 @@ This is a Tensorflow implementation of the Residual Encoder Network based on [Au
 * ![11](images/11.png)
 * ![12](images/12.png)
 
-More examples can be found at [sample_output_images](https://github.com/Armour/Automatic-Image-Colorization/blob/master/sample_output_images) folder.
+* More example output images can be found in [sample_output_images](https://github.com/Armour/Automatic-Image-Colorization/blob/master/sample_output_images) folder.
 
 ## References
 

common.py

@@ -35,64 +35,64 @@ def init_model(train=True):
     :param train: indicate if current is in training
     :return: all stuffs that need for this model
     """
-    # Create training summary folder if not exist.
+    # Create training summary folder if not exist
     create_folder("summary/train/images")
 
-    # Create testing summary folder if not exist.
+    # Create testing summary folder if not exist
     create_folder("summary/test/images")
 
-    # Use gpu if exist.
+    # Use gpu if exist
     with tf.device('/device:GPU:0'):
-        # Init image data file path.
+        # Init image data file path
         print("⏳ Init input file path...")
         if train:
             file_paths = init_file_path(training_dir)
         else:
             file_paths = init_file_path(testing_dir)
 
-        # Init training flag and global step.
+        # Init training flag and global step
         print("⏳ Init placeholder and variables...")
         is_training = tf.placeholder(tf.bool, name="is_training")
         global_step = tf.train.get_or_create_global_step()
 
-        # Load vgg16 model.
+        # Load vgg16 model
         print("🤖 Load vgg16 model...")
         vgg = vgg16.Vgg16()
 
-        # Build residual encoder model.
+        # Build residual encoder model
         print("🤖 Build residual encoder model...")
         residual_encoder = ResidualEncoder()
 
-        # Get dataset iterator.
+        # Get dataset iterator
         iterator = get_dataset_iterator(file_paths, batch_size, shuffle=True)
 
-        # Get color image.
+        # Get color image
         color_image_rgb = iterator.get_next(name="color_image_rgb")
         color_image_yuv = rgb_to_yuv(color_image_rgb, "color_image_yuv")
 
-        # Get gray image.
+        # Get gray image
         gray_image_one_channel = tf.image.rgb_to_grayscale(color_image_rgb, name="gray_image_one_channel")
         gray_image_three_channels = tf.image.grayscale_to_rgb(gray_image_one_channel, name="gray_image_three_channels")
         gray_image_yuv = rgb_to_yuv(gray_image_three_channels, "gray_image_yuv")
 
-        # Build vgg model.
+        # Build vgg model
         with tf.name_scope("vgg16"):
             vgg.build(gray_image_three_channels)
 
-        # Predict model.
+        # Predict model
         predict = residual_encoder.build(input_data=gray_image_three_channels, vgg=vgg, is_training=is_training)
         predict_yuv = tf.concat(axis=3, values=[tf.slice(gray_image_yuv, [0, 0, 0, 0], [-1, -1, -1, 1], name="gray_image_y"), predict], name="predict_yuv")
         predict_rgb = yuv_to_rgb(predict_yuv, "predict_rgb")
 
-        # Get loss.
+        # Get loss
         loss = residual_encoder.get_loss(predict_val=predict, real_val=tf.slice(color_image_yuv, [0, 0, 0, 1], [-1, -1, -1, 2], name="color_image_uv"))
 
-        # Prepare optimizer.
+        # Prepare optimizer
         update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
         with tf.control_dependencies(update_ops):
             optimizer = tf.train.AdamOptimizer().minimize(loss, global_step=global_step, name='optimizer')
 
-        # Init tensorflow summaries.
+        # Init tensorflow summaries
         print("⏳ Init tensorflow summaries...")
         tf.summary.histogram("loss", loss)
         tf.summary.image("gray_image", gray_image_three_channels, max_outputs=1)

config.py

@@ -12,37 +12,37 @@
 import tensorflow as tf
 
 
-# Debug flag, if true, will check model shape using assert in each step and skip gray image check part (to save time).
+# Debug flag, if true, will check model shape using assert in each step and skip gray image check part (to save time)
 debug = False
 
-# Image size for training.
+# Image size for training
 image_size = 224
 
-# Image resize method.
+# Image resize method
 image_resize_method = tf.image.ResizeMethod.BILINEAR
 
-# Parameters for neural network.
-training_iters = 3000000  # The training iterations number.
-batch_size = 6 # Batch size for training data.
-display_step = 50  # Step interval for displaying loss and saving summary during training phase.
-testing_step = 1000  # Step interval for testing and saving image during training phase.
-saving_step = 10000  # Step interval for saving model during training phase.
+# Parameters for neural network
+training_iters = 3000000  # The training iterations number
+batch_size = 6 # Batch size for training data
+display_step = 50  # Step interval for displaying loss and saving summary during training phase
+testing_step = 1000  # Step interval for testing and saving image during training phase
+saving_step = 10000  # Step interval for saving model during training phase
 shuffle_buffer_size = 2000
 
 # UV channel normalization parameters
 u_norm_para = 0.435912
 v_norm_para = 0.614777
 
-# Directory for training and testing dataset.
+# Directory for training and testing dataset
 training_dir = "train2014"
 testing_dir = "test2014"
 
-# Model, result and generated images stored path.
+# Model, result and generated images stored path
 summary_path = "summary"
 training_summary = summary_path + "/train"
 testing_summary = summary_path + "/test"
 
-# Weights for each layer (trainable).
+# Weights for each layer (trainable)
 weights = {
     'b_conv4': tf.Variable(tf.truncated_normal([1, 1, 512, 256], stddev=0.01), trainable=True),
     'b_conv3': tf.Variable(tf.truncated_normal([3, 3, 256, 128], stddev=0.01), trainable=True),
@@ -52,7 +52,7 @@
     'output_conv': tf.Variable(tf.truncated_normal([3, 3, 3, 2], stddev=0.01), trainable=True),
 }
 
-# Gaussian blur kernel (not trainable).
+# Gaussian blur kernel (not trainable)
 gaussin_blur_3x3 = np.divide([
     [1., 2., 1.],
     [2., 4., 2.],

image_helper.py

@@ -21,22 +21,22 @@ def rgb_to_yuv(rgb_image, scope):
     :return: an image with YUV color space
     """
     with tf.name_scope(scope):
-        # Get r, g, b channel.
+        # Get r, g, b channel
         _r = tf.slice(rgb_image, [0, 0, 0, 0], [-1, -1, -1, 1])
         _g = tf.slice(rgb_image, [0, 0, 0, 1], [-1, -1, -1, 1])
         _b = tf.slice(rgb_image, [0, 0, 0, 2], [-1, -1, -1, 1])
 
-        # Calculate y, u, v channel.
+        # Calculate y, u, v channel
         # https://www.pcmag.com/encyclopedia/term/55166/yuv-rgb-conversion-formulas
         _y = 0.299 * _r + 0.587 * _g + 0.114 * _b
         _u = 0.492 * (_b - _y)
         _v = 0.877 * (_r - _y)
 
-        # Normalize u, v channel.
+        # Normalize u, v channel
         _u = _u / (u_norm_para * 2) + 0.5
         _v = _v / (v_norm_para * 2) + 0.5
 
-        # Get image with YUV color space.
+        # Get image with YUV color space
         return tf.clip_by_value(tf.concat(axis=3, values=[_y, _u, _v]), 0.0, 1.0)
 
 def yuv_to_rgb(yuv_image, scope):
@@ -47,22 +47,22 @@ def yuv_to_rgb(yuv_image, scope):
     :return: an image with RGB color space
     """
     with tf.name_scope(scope):
-        # Get y, u, v channel.
+        # Get y, u, v channel
         _y = tf.slice(yuv_image, [0, 0, 0, 0], [-1, -1, -1, 1])
         _u = tf.slice(yuv_image, [0, 0, 0, 1], [-1, -1, -1, 1])
         _v = tf.slice(yuv_image, [0, 0, 0, 2], [-1, -1, -1, 1])
 
-        # Denormalize u, v channel.
+        # Denormalize u, v channel
         _u = (_u - 0.5) * u_norm_para * 2
         _v = (_v - 0.5) * v_norm_para * 2
 
-        # Calculate r, g, b channel.
+        # Calculate r, g, b channel
         # https://www.pcmag.com/encyclopedia/term/55166/yuv-rgb-conversion-formulas
         _r = _y + 1.14 * _v
         _g = _y - 0.395 * _u - 0.581 * _v
         _b = _y + 2.033 * _u
 
-        # Get image with RGB color space.
+        # Get image with RGB color space
         return tf.clip_by_value(tf.concat(axis=3, values=[_r, _g, _b]), 0.0, 1.0)
 
 def concat_images(img_a, img_b):

read_input.py

@@ -28,12 +28,12 @@ def init_file_path(directory):
         print("Throwing all gray space images now... (this will take a long time if the training dataset is huge)")
 
     for file_name in os.listdir(directory):
-        # Skip files that is not jpg.
+        # Skip files that is not jpg
         file_path = '%s/%s' % (directory, file_name)
         if imghdr.what(file_path) is not 'jpeg':
             continue
         if not debug:
-            # Delete all gray space images.
+            # Delete all gray space images
             is_gray_space = True
             img = cv2.imread(file_path, cv2.IMREAD_UNCHANGED)
             if len(img.shape) == 3 and img.shape[2] == 3:
@@ -63,13 +63,13 @@ def read_image(filename):
     :param filename_queue: the filename queue for image files
     :return: image with RGB color space
     """
-    # Read image file.
+    # Read image file
     content = tf.read_file(filename)
-    # Decode the image with RGB color space.
+    # Decode the image with RGB color space
     rgb_image = tf.image.decode_jpeg(content, channels=3, name="color_image_original")
-    # Resize image to the right image_size.
+    # Resize image to the right image_size
     rgb_image = tf.image.resize_images(rgb_image, [image_size, image_size], method=image_resize_method)
-    # Map all pixel element value into [0, 1].
+    # Map all pixel element value into [0, 1]
     return tf.clip_by_value(tf.div(tf.cast(rgb_image, tf.float32), 255), 0.0, 1.0, name="color_image_in_0_1")
 
 

residual_encoder.py

@@ -94,15 +94,15 @@ def build(self, input_data, vgg, is_training):
         if debug:
             assert input_data.get_shape().as_list()[1:] == [224, 224, 3]
 
-        # Batch norm and 1x1 convolutional layer 4.
+        # Batch norm and 1x1 convolutional layer 4
         bn_4 = self.batch_normal(vgg.conv4_3, "bn_4", is_training)
         b_conv4 = self.conv_layer(bn_4, "b_conv4", is_training, bn=False)
 
         if debug:
             assert bn_4.get_shape().as_list()[1:] == [28, 28, 512]
             assert b_conv4.get_shape().as_list()[1:] == [28, 28, 256]
 
-        # Backward upscale layer 4 and add convolutional layer 3.
+        # Backward upscale layer 4 and add convolutional layer 3
         b_conv4_upscale = tf.image.resize_images(b_conv4, [56, 56], method=image_resize_method)
         bn_3 = self.batch_normal(vgg.conv3_3, "bn_3", is_training)
         b_conv3_input = tf.add(bn_3, b_conv4_upscale, name="b_conv3_input")
@@ -114,7 +114,7 @@ def build(self, input_data, vgg, is_training):
             assert b_conv3_input.get_shape().as_list()[1:] == [56, 56, 256]
             assert b_conv3.get_shape().as_list()[1:] == [56, 56, 128]
 
-        # Backward upscale layer 3 and add convolutional layer 2.
+        # Backward upscale layer 3 and add convolutional layer 2
         b_conv3_upscale = tf.image.resize_images(b_conv3, [112, 112], method=image_resize_method)
         bn_2 = self.batch_normal(vgg.conv2_2, "bn_2", is_training)
         b_conv2_input = tf.add(bn_2, b_conv3_upscale, name="b_conv2_input")
@@ -126,7 +126,7 @@ def build(self, input_data, vgg, is_training):
             assert b_conv2_input.get_shape().as_list()[1:] == [112, 112, 128]
             assert b_conv2.get_shape().as_list()[1:] == [112, 112, 64]
 
-        # Backward upscale layer 2 and add convolutional layer 1.
+        # Backward upscale layer 2 and add convolutional layer 1
         b_conv2_upscale = tf.image.resize_images(b_conv2, [224, 224], method=image_resize_method)
         bn_1 = self.batch_normal(vgg.conv1_2, "bn_1", is_training)
         b_conv1_input = tf.add(bn_1, b_conv2_upscale, name="b_conv1_input")
@@ -138,7 +138,7 @@ def build(self, input_data, vgg, is_training):
             assert b_conv1_input.get_shape().as_list()[1:] == [224, 224, 64]
             assert b_conv1.get_shape().as_list()[1:] == [224, 224, 3]
 
-        # Backward upscale layer 1 and add input layer.
+        # Backward upscale layer 1 and add input layer
         bn_0 = self.batch_normal(input_data, "bn_0", is_training)
         b_conv0_input = tf.add(bn_0, b_conv1, name="b_conv0_input")
         b_conv0 = self.conv_layer(b_conv0_input, "b_conv0", is_training)
@@ -148,7 +148,7 @@ def build(self, input_data, vgg, is_training):
             assert b_conv0_input.get_shape().as_list()[1:] == [224, 224, 3]
             assert b_conv0.get_shape().as_list()[1:] == [224, 224, 3]
 
-        # Output layer.
+        # Output layer
         output_layer = self.conv_layer(b_conv0, "output_conv", is_training, relu=False)
 
         if debug:

train.py

@@ -18,32 +18,32 @@
 
 
 if __name__ == '__main__':
-    # Init model.
+    # Init model
     is_training, global_step, optimizer, loss, predict_rgb, color_image_rgb, gray_image, _ = init_model(train=True)
 
-    # Init scaffold, hooks and config.
+    # Init scaffold, hooks and config
     scaffold = tf.train.Scaffold()
     summary_hook = tf.train.SummarySaverHook(output_dir=training_summary, save_steps=display_step, scaffold=scaffold)
     checkpoint_hook = tf.train.CheckpointSaverHook(checkpoint_dir=summary_path, save_steps=saving_step, scaffold=scaffold)
     num_step_hook = tf.train.StopAtStepHook(num_steps=training_iters)
     config = tf.ConfigProto(allow_soft_placement=True, log_device_placement=True, gpu_options=(tf.GPUOptions(allow_growth=True)))
 
-    # Create a session for running operations in the Graph.
+    # Create a session for running operations in the Graph
     with tf.train.MonitoredTrainingSession(checkpoint_dir=summary_path,
                                            hooks=[summary_hook, checkpoint_hook, num_step_hook],
                                            scaffold=scaffold,
                                            config=config) as sess:
         print("🤖 Start training...")
 
         while not sess.should_stop():
-            # Run optimizer.
+            # Run optimizer
             _, step, l, pred, color, gray = sess.run([optimizer, global_step, loss, predict_rgb, color_image_rgb, gray_image] , feed_dict={is_training: True})
 
             if step % display_step == 0:
-                # Print batch loss.
+                # Print batch loss
                 print("📖 Iter %d, Minibatch Loss = %f" % (step, l))
 
-                # Save testing image.
+                # Save testing image
                 if step % testing_step == 0:
                     summary_image = concat_images(gray[0], pred[0])
                     summary_image = concat_images(summary_image, color[0])