Refactored code

INetwork.py

@@ -34,45 +34,73 @@
 parser = argparse.ArgumentParser(description='Neural style transfer with Keras.')
 parser.add_argument('base_image_path', metavar='base', type=str,
                     help='Path to the image to transform.')
+
 parser.add_argument('syle_image_paths', metavar='ref', nargs='+', type=str,
                     help='Path to the style reference image.')
+
 parser.add_argument('result_prefix', metavar='res_prefix', type=str,
                     help='Prefix for the saved results.')
-parser.add_argument("--style_masks", type=str, default=None, nargs='+', help='Masks for style images')
-parser.add_argument("--color_mask", type=str, default=None, help='Mask for color preservation')
 
-parser.add_argument("--image_size", dest="img_size", default=400, type=int, help='Output Image size')
+parser.add_argument("--style_masks", type=str, default=None, nargs='+',
+                    help='Masks for style images')
+
+parser.add_argument("--color_mask", type=str, default=None,
+                    help='Mask for color preservation')
+
+parser.add_argument("--image_size", dest="img_size", default=400, type=int,
+                    help='Minimum image size')
+
 parser.add_argument("--content_weight", dest="content_weight", default=0.025, type=float,
-                    help="Weight of content")  # 0.025
-parser.add_argument("--style_weight", dest="style_weight", nargs='+', default=[1], type=float, help="Weight of content")  # 1.0
+                    help="Weight of content")
+
+parser.add_argument("--style_weight", dest="style_weight", nargs='+', default=[1], type=float,
+                    help="Weight of style, can be multiple for multiple styles")
+
 parser.add_argument("--style_scale", dest="style_scale", default=1.0, type=float,
-                    help="Scale the weightage of the style")  # 1, 0.5, 2
+                    help="Scale the weighing of the style")
+
 parser.add_argument("--total_variation_weight", dest="tv_weight", default=8.5e-5, type=float,
-                    help="Total Variation in the Weights")  # 1.0
+                    help="Total Variation weight")
+
+parser.add_argument("--num_iter", dest="num_iter", default=10, type=int,
+                    help="Number of iterations")
 
-parser.add_argument("--num_iter", dest="num_iter", default=10, type=int, help="Number of iterations")
-parser.add_argument("--model", default="vgg16", type=str, help="Choices are 'vgg16' and 'vgg19'")
-parser.add_argument("--content_loss_type", default=0, type=int, help='Can be one of 0, 1 or 2. Readme contains '
-                                                                     'the required information of each mode.')
+parser.add_argument("--model", default="vgg16", type=str,
+                    help="Choices are 'vgg16' and 'vgg19'")
+
+parser.add_argument("--content_loss_type", default=0, type=int,
+                    help='Can be one of 0, 1 or 2. Readme contains the required information of each mode.')
 
 parser.add_argument("--rescale_image", dest="rescale_image", default="False", type=str,
                     help="Rescale image after execution to original dimentions")
+
 parser.add_argument("--rescale_method", dest="rescale_method", default="bilinear", type=str,
                     help="Rescale image algorithm")
+
 parser.add_argument("--maintain_aspect_ratio", dest="maintain_aspect_ratio", default="True", type=str,
-                    help="Maintain aspect ratio of image")
+                    help="Maintain aspect ratio of loaded images")
+
+parser.add_argument("--content_layer", dest="content_layer", default="conv5_2", type=str,
+                    help="Content layer used for content loss.")
 
-parser.add_argument("--content_layer", dest="content_layer", default="conv5_2", type=str, help="Optional 'conv4_2'")
 parser.add_argument("--init_image", dest="init_image", default="content", type=str,
                     help="Initial image used to generate the final image. Options are 'content', 'noise', or 'gray'")
+
 parser.add_argument("--pool_type", dest="pool", default="max", type=str,
-                    help='Pooling type. Can be "ave" for average pooling'
-                         ' or "max" for max pooling ')
+                    help='Pooling type. Can be "ave" for average pooling or "max" for max pooling')
+
 parser.add_argument('--preserve_color', dest='color', default="False", type=str,
                     help='Preserve original color in image')
+
 parser.add_argument('--min_improvement', default=0.0, type=float,
                     help='Defines minimum improvement required to continue script')
 
+
+def str_to_bool(v):
+    return v.lower() in ("true", "yes", "t", "1")
+
+''' Arguments '''
+
 args = parser.parse_args()
 base_image_path = args.base_image_path
 style_reference_image_paths = args.syle_image_paths
@@ -91,17 +119,12 @@
 
 if style_masks_present:
     assert len(style_image_paths) == len(mask_paths), "Wrong number of style masks provided.\n" \
-                                                               "Number of style images = %d, \n" \
-                                                               "Number of style mask paths = %d." % \
-                                                               (len(style_image_paths), len(style_masks_present))
+                                                      "Number of style images = %d, \n" \
+                                                      "Number of style mask paths = %d." % \
+                                                      (len(style_image_paths), len(style_masks_present))
 
 color_mask_present = args.color_mask is not None
 
-
-def str_to_bool(v):
-    return v.lower() in ("true", "yes", "t", "1")
-
-
 rescale_image = str_to_bool(args.rescale_image)
 maintain_aspect_ratio = str_to_bool(args.maintain_aspect_ratio)
 preserve_color = str_to_bool(args.color)
@@ -126,6 +149,14 @@ def str_to_bool(v):
     for style_weight in args.style_weight:
         style_weights.append(style_weight * args.style_scale)
 
+# Decide pooling function
+pooltype = str(args.pool).lower()
+assert pooltype in ["ave", "max"], 'Pooling argument is wrong. Needs to be either "ave" or "max".'
+
+pooltype = 1 if pooltype == "ave" else 0
+
+read_mode = "gray" if args.init_image == "gray" else "color"
+
 # dimensions of the generated picture.
 img_width = img_height = 0
 
@@ -135,6 +166,7 @@ def str_to_bool(v):
 assert args.init_image in ["content", "noise", "gray"], "init_image must be one of ['content', 'noise', 'gray']"
 assert args.content_loss_type in [0, 1, 2], "Content Loss Type must be one of 0, 1 or 2"
 
+
 # util function to open, resize and format pictures into appropriate tensors
 def preprocess_image(image_path, load_dims=False, read_mode="color"):
     global img_width, img_height, img_WIDTH, img_HEIGHT, aspect_ratio
@@ -246,21 +278,13 @@ def load_mask(mask_path, shape, return_mask_img=False):
     return mask_tensor
 
 
-# Decide pooling function
-pooltype = str(args.pool).lower()
-assert pooltype in ["ave", "max"], 'Pooling argument is wrong. Needs to be either "ave" or "max".'
-
-pooltype = 1 if pooltype == "ave" else 0
-
-
 def pooling_func(x):
     if pooltype == 1:
         return AveragePooling2D((2, 2), strides=(2, 2))(x)
     else:
         return MaxPooling2D((2, 2), strides=(2, 2))(x)
 
 
-read_mode = "gray" if args.init_image == "gray" else "color"
 # get tensor representations of our images
 base_image = K.variable(preprocess_image(base_image_path, True, read_mode=read_mode))
 

Network.py

@@ -32,45 +32,73 @@
 parser = argparse.ArgumentParser(description='Neural style transfer with Keras.')
 parser.add_argument('base_image_path', metavar='base', type=str,
                     help='Path to the image to transform.')
+
 parser.add_argument('syle_image_paths', metavar='ref', nargs='+', type=str,
                     help='Path to the style reference image.')
+
 parser.add_argument('result_prefix', metavar='res_prefix', type=str,
                     help='Prefix for the saved results.')
-parser.add_argument("--style_masks", type=str, default=None, nargs='+', help='Masks for style images')
-parser.add_argument("--color_mask", type=str, default=None, help='Mask for color preservation')
 
-parser.add_argument("--image_size", dest="img_size", default=400, type=int, help='Output Image size')
+parser.add_argument("--style_masks", type=str, default=None, nargs='+',
+                    help='Masks for style images')
+
+parser.add_argument("--color_mask", type=str, default=None,
+                    help='Mask for color preservation')
+
+parser.add_argument("--image_size", dest="img_size", default=400, type=int,
+                    help='Minimum image size')
+
 parser.add_argument("--content_weight", dest="content_weight", default=0.025, type=float,
-                    help="Weight of content")  # 0.025
-parser.add_argument("--style_weight", dest="style_weight", nargs='+', default=[1], type=float, help="Weight of content")  # 1.0
+                    help="Weight of content")
+
+parser.add_argument("--style_weight", dest="style_weight", nargs='+', default=[1], type=float,
+                    help="Weight of style, can be multiple for multiple styles")
+
 parser.add_argument("--style_scale", dest="style_scale", default=1.0, type=float,
-                    help="Scale the weightage of the style")  # 1, 0.5, 2
+                    help="Scale the weighing of the style")
+
 parser.add_argument("--total_variation_weight", dest="tv_weight", default=8.5e-5, type=float,
-                    help="Total Variation in the Weights")  # 1.0
+                    help="Total Variation weight")
+
+parser.add_argument("--num_iter", dest="num_iter", default=10, type=int,
+                    help="Number of iterations")
 
-parser.add_argument("--num_iter", dest="num_iter", default=10, type=int, help="Number of iterations")
-parser.add_argument("--model", default="vgg16", type=str, help="Choices are 'vgg16' and 'vgg19'")
-parser.add_argument("--content_loss_type", default=0, type=int, help='Can be one of 0, 1 or 2. Readme contains '
-                                                                     'the required information of each mode.')
+parser.add_argument("--model", default="vgg16", type=str,
+                    help="Choices are 'vgg16' and 'vgg19'")
+
+parser.add_argument("--content_loss_type", default=0, type=int,
+                    help='Can be one of 0, 1 or 2. Readme contains the required information of each mode.')
 
 parser.add_argument("--rescale_image", dest="rescale_image", default="False", type=str,
                     help="Rescale image after execution to original dimentions")
+
 parser.add_argument("--rescale_method", dest="rescale_method", default="bilinear", type=str,
                     help="Rescale image algorithm")
+
 parser.add_argument("--maintain_aspect_ratio", dest="maintain_aspect_ratio", default="True", type=str,
-                    help="Maintain aspect ratio of image")
+                    help="Maintain aspect ratio of loaded images")
+
+parser.add_argument("--content_layer", dest="content_layer", default="conv5_2", type=str,
+                    help="Content layer used for content loss.")
 
-parser.add_argument("--content_layer", dest="content_layer", default="conv5_2", type=str, help="Optional 'conv4_2'")
 parser.add_argument("--init_image", dest="init_image", default="content", type=str,
                     help="Initial image used to generate the final image. Options are 'content', 'noise', or 'gray'")
+
 parser.add_argument("--pool_type", dest="pool", default="max", type=str,
-                    help='Pooling type. Can be "ave" for average pooling'
-                         ' or "max" for max pooling ')
+                    help='Pooling type. Can be "ave" for average pooling or "max" for max pooling')
+
 parser.add_argument('--preserve_color', dest='color', default="False", type=str,
                     help='Preserve original color in image')
+
 parser.add_argument('--min_improvement', default=0.0, type=float,
                     help='Defines minimum improvement required to continue script')
 
+
+def str_to_bool(v):
+    return v.lower() in ("true", "yes", "t", "1")
+
+''' Arguments '''
+
 args = parser.parse_args()
 base_image_path = args.base_image_path
 style_reference_image_paths = args.syle_image_paths
@@ -89,17 +117,12 @@
 
 if style_masks_present:
     assert len(style_image_paths) == len(mask_paths), "Wrong number of style masks provided.\n" \
-                                                               "Number of style images = %d, \n" \
-                                                               "Number of style mask paths = %d." % \
-                                                               (len(style_image_paths), len(style_masks_present))
+                                                      "Number of style images = %d, \n" \
+                                                      "Number of style mask paths = %d." % \
+                                                      (len(style_image_paths), len(style_masks_present))
 
 color_mask_present = args.color_mask is not None
 
-
-def str_to_bool(v):
-    return v.lower() in ("true", "yes", "t", "1")
-
-
 rescale_image = str_to_bool(args.rescale_image)
 maintain_aspect_ratio = str_to_bool(args.maintain_aspect_ratio)
 preserve_color = str_to_bool(args.color)
@@ -124,6 +147,14 @@ def str_to_bool(v):
     for style_weight in args.style_weight:
         style_weights.append(style_weight * args.style_scale)
 
+# Decide pooling function
+pooltype = str(args.pool).lower()
+assert pooltype in ["ave", "max"], 'Pooling argument is wrong. Needs to be either "ave" or "max".'
+
+pooltype = 1 if pooltype == "ave" else 0
+
+read_mode = "gray" if args.init_image == "gray" else "color"
+
 # dimensions of the generated picture.
 img_width = img_height = 0
 
@@ -133,6 +164,7 @@ def str_to_bool(v):
 assert args.init_image in ["content", "noise", "gray"], "init_image must be one of ['content', 'noise', 'gray']"
 assert args.content_loss_type in [0, 1, 2], "Content Loss Type must be one of 0, 1 or 2"
 
+
 # util function to open, resize and format pictures into appropriate tensors
 def preprocess_image(image_path, load_dims=False, read_mode="color"):
     global img_width, img_height, img_WIDTH, img_HEIGHT, aspect_ratio
@@ -244,21 +276,13 @@ def load_mask(mask_path, shape, return_mask_img=False):
     return mask_tensor
 
 
-# Decide pooling function
-pooltype = str(args.pool).lower()
-assert pooltype in ["ave", "max"], 'Pooling argument is wrong. Needs to be either "ave" or "max".'
-
-pooltype = 1 if pooltype == "ave" else 0
-
-
 def pooling_func(x):
     if pooltype == 1:
         return AveragePooling2D((2, 2), strides=(2, 2))(x)
     else:
         return MaxPooling2D((2, 2), strides=(2, 2))(x)
 
 
-read_mode = "gray" if args.init_image == "gray" else "color"
 # get tensor representations of our images
 base_image = K.variable(preprocess_image(base_image_path, True, read_mode=read_mode))