Add: Source

src/config.py

@@ -0,0 +1,15 @@
+import os
+
+DATASET_URL = 'https://download.microsoft.com/download/3/E/1/3E1C3F21-ECDB-4869-8368-6DEBA77B919F/kagglecatsanddogs_3367a.zip'
+
+CACHE_DIR = os.getcwd()
+CACHE_SUBDIR = 'data'
+
+if not os.path.isdir(CACHE_SUBDIR):
+    os.mkdir(CACHE_SUBDIR)
+
+DATASET_PATH = os.path.join(CACHE_DIR, CACHE_SUBDIR, 'PetImages')
+
+IMAGE_SIZE = (180, 180)
+BATCH_SIZE = 32
+EPOCHS = 50

src/draw.py

@@ -0,0 +1,21 @@
+from typing import Optional
+
+import tensorflow as tf
+import matplotlib.pyplot as plt
+
+
+def visualize_data(dataset: tf.data.Dataset, data_augmentation: Optional[tf.keras.Sequential]=None) -> None:
+    plt.figure(figsize=(10, 10))
+    for images, labels in dataset.take(1):
+        for i in range(9):
+            _ = plt.subplot(3, 3, i + 1)
+            
+            if data_augmentation != None:
+                augmented_image = data_augmentation(images)
+                plt.imshow(augmented_image[0].numpy().astype('uint8'))
+            else:
+                plt.imshow(images[i].numpy().astype('uint8'))
+            
+            plt.title(int(labels[i]))
+            plt.axis('off')
+    plt.show()

src/models.py

@@ -0,0 +1,77 @@
+from typing import Tuple
+from typing import Optional
+
+import tensorflow as tf
+from tensorflow.keras import layers
+from keras_tuner import HyperModel
+
+class MakeHyperModel(HyperModel):
+    def __init__(self, input_shape: Tuple[int, int, int], num_classes: int, data_augmentation: Optional[tf.keras.Sequential] = None) -> None:
+        self.input_shape = input_shape
+        self.num_classes = num_classes
+        self.data_augmentation = data_augmentation
+    
+    def build(self, hp) -> tf.keras.Model:
+        inputs = tf.keras.Input(shape=self.input_shape)
+        
+        if self.data_augmentation != None:
+            x = self.data_augmentation(inputs)
+        else:
+            x = inputs
+        
+        x = layers.Rescaling(1.0/255)(x)
+        x = layers.Conv2D(32, 3, strides=2, padding='same')(x)
+        x = layers.BatchNormalization()(x)
+        x = layers.Activation('relu')(x)
+        
+        x = layers.Conv2D(64, 3,  padding='same')(x)
+        x = layers.BatchNormalization()(x)
+        x = layers.Activation('relu')(x)
+        
+        previous_block_activation = x
+        
+        for size in [128, 256, 512, 728]:
+            x = layers.Activation('relu')(x)
+            x = layers.SeparableConv2D(size, 3, padding='same')(x)
+            x = layers.BatchNormalization()(x)
+            
+            x = layers.Activation("relu")(x)
+            x = layers.SeparableConv2D(size, 3, padding='same')(x)
+            x = layers.BatchNormalization()(x)
+            
+            x = layers.MaxPooling2D(3, strides=2, padding='same')(x)
+            
+            residual = layers.Conv2D(size, 1, strides=2, padding='same')(previous_block_activation)
+            
+            x = layers.add([x, residual])
+            previous_block_activation = x
+        
+        x = layers.SeparableConv2D(1024, 3, padding='same')(x)
+        x = layers.BatchNormalization()(x)
+        x = layers.Activation('relu')(x)
+        
+        x = layers.GlobalAveragePooling2D()(x)
+        
+        if self.num_classes == 2:
+            activation = 'sigmoid'
+            loss_fn = 'binary_crossentropy'
+            units = 1
+        else:
+            activation = 'softmax'
+            loss_fn = 'categorical_crossentropy'
+            units = self.num_classes
+        
+        x = layers.Dropout(0.5)(x)
+        outputs = layers.Dense(units, activation=activation)(x)
+        
+        model = tf.keras.Model(inputs, outputs)
+        
+        model.compile(
+            optimizer=tf.keras.optimizers.Adam(
+                hp.Choice("learning_rate", values=[1e-2, 1e-3, 1e-4])
+            ),
+            loss=loss_fn,
+            metrics=['accuracy']
+        )
+        
+        return model

src/preprocessing.py

@@ -0,0 +1,29 @@
+import os
+from typing import Tuple
+
+import tensorflow as tf
+
+def delete_corrupted_image(dataset_path: str, categories: Tuple[str]) -> int:
+    num_skipped = 0
+    
+    for folder_name in categories:
+        folder_path = os.path.join(dataset_path, folder_name)
+        for fname in os.listdir(folder_path):
+            fpath = os.path.join(folder_path, fname)
+            try:
+                fobj = open(fpath, 'rb')
+                is_jfif = tf.compat.as_bytes("JFIF") in fobj.peek(10)
+            finally:
+                fobj.close()
+
+            if not is_jfif:
+                num_skipped += 1
+                os.remove(fpath)
+            
+    return num_skipped
+
+def get_data_augmentation() -> tf.keras.Sequential:
+    return tf.keras.Sequential([
+        tf.keras.layers.RandomFlip('horizontal'),
+        tf.keras.layers.RandomRotation(0.1)
+    ])