Update Deblurring model_modded.py

Deblurring model_modded.py

@@ -8,7 +8,6 @@ from tensorflow.keras.models import Model
 from tensorflow.keras.optimizers import Adam
 import cv2
 
-# Residual Block for the Generator
 def residual_block(x, filters):
     res = Conv2D(filters, kernel_size=3, strides=1, padding='same',
                  kernel_initializer=tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02))(x)
@@ -19,7 +18,6 @@ def residual_block(x, filters):
     res = InstanceNormalization()(res)
     return Add()([x, res])
 
-# Generator Architecture
 def build_generator():
     inputs = Input(shape=(256, 256, 3))
     x = Conv2D(64, kernel_size=7, strides=1, padding='same',
@@ -27,7 +25,6 @@ def build_generator():
     x = InstanceNormalization()(x)
     x = Activation('relu')(x)
 
-    # Down-sampling
     x = Conv2D(128, kernel_size=3, strides=2, padding='same',
                kernel_initializer=tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02))(x)
     x = InstanceNormalization()(x)
@@ -38,11 +35,9 @@ def build_generator():
     x = InstanceNormalization()(x)
     x = Activation('relu')(x)
 
-    # Residual Blocks
     for _ in range(9):
         x = residual_block(x, 256)
 
-    # Up-sampling
     x = Conv2DTranspose(128, kernel_size=3, strides=2, padding='same',
                         kernel_initializer=tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02))(x)
     x = InstanceNormalization()(x)
@@ -57,7 +52,6 @@ def build_generator():
                      kernel_initializer=tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.02))(x)
     return Model(inputs, outputs, name="Generator")
 
-# Discriminator Architecture
 def build_discriminator():
     inputs = Input(shape=(256, 256, 3))
 
@@ -81,14 +75,12 @@ def build_discriminator():
     x = LeakyReLU(0.2)(x)
 
     x = GlobalAveragePooling2D()(x)
-    outputs = Dense(1)(x)  # Wasserstein loss raw output
+    outputs = Dense(1)(x) 
     return Model(inputs, outputs, name="Discriminator")
 
-# Wasserstein Loss
 def wasserstein_loss(y_true, y_pred):
     return -tf.reduce_mean(y_true * y_pred)
 
-# Load and Preprocess Data
 def load_images(folder):
     images = []
     for filename in os.listdir(folder):
@@ -98,7 +90,6 @@ def load_images(folder):
             images.append(img)
     return np.array(images)
 
-# Training Pipeline
 def train(generator, discriminator, blurred_images, clear_images, epochs, batch_size):
     optimizer_g = Adam(learning_rate=0.0001, beta_1=0.5, beta_2=0.999)
     optimizer_d = Adam(learning_rate=0.0002, beta_1=0.5, beta_2=0.999)
@@ -106,38 +97,31 @@ def train(generator, discriminator, blurred_images, clear_images, epochs, batch_
     for epoch in range(epochs):
         print(f"Epoch {epoch + 1}/{epochs}")
         for i in range(0, len(blurred_images), batch_size):
-            # Prepare batches
             blurred_batch = blurred_images[i:i + batch_size]
             clear_batch = clear_images[i:i + batch_size]
 
-            # Generate fake images
             fake_images = generator.predict(blurred_batch)
 
-            # Discriminator Training
-            real_labels = -np.ones((len(clear_batch), 1))  # Label real as -1
-            fake_labels = np.ones((len(fake_images), 1))  # Label fake as 1
+            real_labels = -np.ones((len(clear_batch), 1))  
+            fake_labels = np.ones((len(fake_images), 1))  
             d_loss_real = discriminator.train_on_batch(clear_batch, real_labels)
             d_loss_fake = discriminator.train_on_batch(fake_images, fake_labels)
             d_loss = d_loss_real + d_loss_fake
 
-            # Generator Training
-            misleading_labels = -np.ones((len(blurred_batch), 1))  # Fool discriminator
+            misleading_labels = -np.ones((len(blurred_batch), 1))  
             g_loss = generator.train_on_batch(blurred_batch, misleading_labels)
 
             print(f"Batch {i // batch_size + 1}: D Loss: {d_loss:.4f}, G Loss: {g_loss:.4f}")
 
-# Paths and Data Loading
 blurred_folder = "blurred_sketches"
 clear_folder = "clear_sketches"
 blurred_images = load_images(blurred_folder)
 clear_images = load_images(clear_folder)
 
-# Instantiate Models
 generator = build_generator()
 discriminator = build_discriminator()
 
-# Compile Discriminator
 discriminator.compile(optimizer=Adam(0.0002, 0.5, 0.999), loss=wasserstein_loss)
 
-# Train GAN
+
 train(generator, discriminator, blurred_images, clear_images, epochs=500, batch_size=16)