refactor for image residual

data.py

@@ -1,9 +1,10 @@
 import tensorflow as tf
 import util
 
-resize_func = None
+resize = False
+residual = False
 
-def dataset(hr_flist, lr_flist, scale, resize_func=None):
+def dataset(hr_flist, lr_flist, scale, resize=resize, residual=residual):
     with open(hr_flist) as f:
         hr_filename_list = f.read().splitlines()
     with open(lr_flist) as f:
@@ -15,19 +16,28 @@ def dataset(hr_flist, lr_flist, scale, resize_func=None):
     lr_image = tf.image.decode_image(lr_image_file, channels=3)
     hr_image = tf.image.convert_image_dtype(hr_image, tf.float32)
     lr_image = tf.image.convert_image_dtype(lr_image, tf.float32)
-    hr_patches0, lr_patches0 = make_patches(hr_image, lr_image, scale, resize_func)
-    hr_patches1, lr_patches1 = make_patches(tf.image.rot90(hr_image), tf.image.rot90(lr_image), scale, resize_func)
+    if (residual):
+        hr_image = make_residual(hr_image, lr_image)
+    hr_patches0, lr_patches0 = make_patches(hr_image, lr_image, scale, resize)
+    hr_patches1, lr_patches1 = make_patches(tf.image.rot90(hr_image), tf.image.rot90(lr_image), scale, resize)
     return tf.concat([hr_patches0, hr_patches1], 0), tf.concat([lr_patches0, lr_patches1], 0)
 
-def make_patches(hr_image, lr_image, scale, resize_func):
+def make_residual(hr_image, lr_image):
+    hr_image = tf.expand_dims(hr_image, 0)
+    lr_image = tf.expand_dims(lr_image, 0)
+    hr_image_shape = tf.shape(hr_image)[1:3]
+    res_image = hr_image - util.resize_func(lr_image, hr_image_shape)
+    return tf.reshape(res_image, [hr_image_shape[0], hr_image_shape[1], 3])
+
+def make_patches(hr_image, lr_image, scale, resize):
     hr_image = tf.stack(flip([hr_image]))
     lr_image = tf.stack(flip([lr_image]))    
     hr_patches = util.image_to_patches(hr_image)
-    if (resize_func is None):
-        lr_patches = util.image_to_patches(lr_image, scale)
-    else:
-        lr_image = resize_func(lr_image, tf.shape(hr_image)[1:3])
+    if (resize):
+        lr_image = util.resize_func(lr_image, tf.shape(hr_image)[1:3])
         lr_patches = util.image_to_patches(lr_image)
+    else:
+        lr_patches = util.image_to_patches(lr_image, scale)
     return hr_patches, lr_patches 
 
 def flip(img_list):

data_residual.py

@@ -0,0 +1,8 @@
+import tensorflow as tf
+import data
+
+resize = False
+residual = True
+
+def dataset(hr_flist, lr_flist, scale):
+    return data.dataset(hr_flist, lr_flist, scale, resize, residual)

data_resize_residual.py

@@ -0,0 +1,8 @@
+import tensorflow as tf
+import data
+
+resize = True
+residual = True
+
+def dataset(hr_flist, lr_flist, scale):
+    return data.dataset(hr_flist, lr_flist, scale, resize, residual)

predict.py

@@ -4,7 +4,7 @@
 flags = tf.app.flags
 FLAGS = flags.FLAGS
 
-flags.DEFINE_string('data_name', 'data_resize', 'Directory to put the training data.')
+flags.DEFINE_string('data_name', 'data_resize_residual', 'Directory to put the training data.')
 flags.DEFINE_string('hr_flist', 'flist/set5_predict.flist', 'file_list put the training data.')
 flags.DEFINE_string('lr_flist', 'flist/set5_lrX2.flist', 'Directory to put the training data.')
 flags.DEFINE_integer('scale', '2', 'batch size for training')
@@ -13,7 +13,7 @@
 
 data = __import__(FLAGS.data_name)
 model = __import__(FLAGS.model_name)
-if ((data.resize_func is None) != model.upsample):
+if (data.resize == model.upsample):
     print "Config Error"
     quit()
 
@@ -30,14 +30,19 @@
     lr_image = tf.expand_dims(lr_image, 0)
     lr_image_shape = tf.shape(lr_image)[1:3]
     hr_image_shape = lr_image_shape * FLAGS.scale
-    if (data.resize_func is not None):
-        lr_image = data.resize_func(lr_image, hr_image_shape)
+    if (data.resize):
+        lr_image = util.resize_func(lr_image, hr_image_shape)
         lr_image = tf.reshape(lr_image, [1, hr_image_shape[0], hr_image_shape[1], 3])
     else:
         lr_image = tf.reshape(lr_image, [1, lr_image_shape[0], lr_image_shape[1], 3])
-    lr_image = util.pad_boundary(lr_image)
-    hr_image = model.build_model(lr_image, FLAGS.scale, training=False, reuse=False)
+    lr_image_padded = util.pad_boundary(lr_image)
+    hr_image = model.build_model(lr_image_padded, FLAGS.scale, training=False, reuse=False)
     hr_image = util.crop_center(hr_image, hr_image_shape)
+    if (data.residual):
+        if (data.resize):
+            hr_image += lr_image
+        else:
+            hr_image += util.resize_func(lr_image, hr_image_shape)
     hr_image = tf.image.convert_image_dtype(hr_image, tf.uint8, saturate=True)
     hr_image = tf.reshape(hr_image, [hr_image_shape[0], hr_image_shape[1], 3])
     hr_image = tf.image.encode_png(hr_image)

run.sh

@@ -25,7 +25,7 @@ set -x
 EXPR_NAME="try"
 TRAIN_DIR="tmp"
 MODEL_NAME="model_resnet"
-DATA_NAME="data_resize"
+DATA_NAME="data_resize_residual"
 HR_FLIST="flist/hr.flist"
 LR_FLIST="flist/lrX2.flist"
 SCALE=2

train.py

@@ -5,7 +5,7 @@
 flags = tf.app.flags
 FLAGS = flags.FLAGS
 
-flags.DEFINE_string('data_name', 'data_resize', 'Directory to put the training data.')
+flags.DEFINE_string('data_name', 'data_resize_residual', 'Directory to put the training data.')
 flags.DEFINE_string('hr_flist', 'flist/hr_debug.flist', 'file_list put the training data.')
 flags.DEFINE_string('lr_flist', 'flist/lrX2_debug.flist', 'Directory to put the training data.')
 flags.DEFINE_integer('scale', '2', 'batch size for training')
@@ -17,7 +17,7 @@
 
 data = __import__(FLAGS.data_name)
 model = __import__(FLAGS.model_name)
-if ((data.resize_func is None) != model.upsample):
+if (data.resize == model.upsample):
     print "Config Error"
     quit()
 

util.py

@@ -1,5 +1,7 @@
 import tensorflow as tf
 
+resize_func = tf.image.resize_nearest_neighbor
+
 def image_to_patches(image, scale=1):
     patch_height = 108 / scale
     patch_width = 108 / scale

validate.py

@@ -4,7 +4,7 @@
 flags = tf.app.flags
 FLAGS = flags.FLAGS
 
-flags.DEFINE_string('data_name', 'data_resize', 'Directory to put the training data.')
+flags.DEFINE_string('data_name', 'data_resize_residual', 'Directory to put the training data.')
 flags.DEFINE_string('hr_flist', 'flist/set5_hr.flist', 'file_list put the training data.')
 flags.DEFINE_string('lr_flist', 'flist/set5_lrX2.flist', 'Directory to put the training data.')
 flags.DEFINE_integer('scale', '2', 'batch size for training')
@@ -13,7 +13,7 @@
 
 data = __import__(FLAGS.data_name)
 model = __import__(FLAGS.model_name)
-if ((data.resize_func is None) != model.upsample):
+if (data.resize == model.upsample):
     print "Config Error"
     quit()
 
@@ -33,15 +33,20 @@
     lr_image = tf.expand_dims(lr_image, 0)
     lr_image_shape = tf.shape(lr_image)[1:3]
     hr_image_shape = tf.shape(hr_image)[1:3]
-    if (data.resize_func is not None):
-        lr_image = data.resize_func(lr_image, hr_image_shape)
+    if (data.resize):
+        lr_image = util.resize_func(lr_image, hr_image_shape)
         lr_image = tf.reshape(lr_image, [1, hr_image_shape[0], hr_image_shape[1], 3])
     else:
         lr_image = tf.reshape(lr_image, [1, lr_image_shape[0], lr_image_shape[1], 3])
-    lr_image = util.pad_boundary(lr_image)
-    lr_image = model.build_model(lr_image, FLAGS.scale, training=False, reuse=False)
-    lr_image = util.crop_center(lr_image, hr_image_shape)
-    error = tf.losses.mean_squared_error(hr_image, lr_image)
+    lr_image_padded = util.pad_boundary(lr_image)
+    sr_image = model.build_model(lr_image_padded, FLAGS.scale, training=False, reuse=False)
+    sr_image = util.crop_center(sr_image, hr_image_shape)
+    if (data.residual):
+        if (data.resize):
+            sr_image += lr_image
+        else:
+            sr_image += util.resize_func(lr_image, hr_image_shape)
+    error = tf.losses.mean_squared_error(hr_image, sr_image)
 
     init = tf.global_variables_initializer()
     init_local = tf.local_variables_initializer()