Upload 2 files

RESNET_images.py

@@ -0,0 +1,272 @@
+# -*- coding: utf-8 -*-
+#images- 2800 training, test- 280
+# 10 epoch, batch= 32, kfold= 4
+# test accuracy 0.75
+# test loss 1.26
+# avarage accuracy 61.84
+#Classification Summary:
+#Real images correctly classified: 76
+#Real images incorrectly classified: 63
+#Fake images correctly classified: 135
+#Fake images incorrectly classified: 5
+###########resnet- מקפיא חלק מהשכבות ההתחלתיות כלומר יש אימון על שכבות רבות##########
+#resnet_model = tf.keras.applications.ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+# Freeze initial layers (optional)
+#for layer in resnet_model.layers[:17]:<<<<=
+#    layer.trainable = False
+# Modify final layer
+#x = resnet_model.output
+#x = tf.keras.layers.Flatten()(x)
+#x = tf.keras.layers.Dense(64, activation='relu')(x)<<<<<<=
+#predictions = tf.keras.layers.Dense(1, activation='sigmoid')(x)  # Binary classification
+
+
+
+##images- 2800 training, test- 280
+#KFold(n_splits=4, batch_size = 32, epochs = 5
+#train only last layer
+#Classification Summary:
+#Average accuracy: 76.95%
+#Test Loss: 0.4741879999637604, Test Accuracy: 0.781361997127533
+#Real images correctly classified: 116
+#Real images incorrectly classified: 23
+#Fake images correctly classified: 102
+#Fake images incorrectly classified: 38
+
+#############
+#saves weight
+##images- 2800 training, test- 280
+#KFold(n_splits=4, batch_size = 32, epochs = 5
+#train only last layer
+#Classification Summary:
+#Average accuracy: 77.11%
+#Test Loss: 0.47622182965278625, Test Accuracy: 0.7777777910232544
+#Classification Summary:
+#Real images correctly classified: 116
+#Real images incorrectly classified: 23
+#Fake images correctly classified: 101
+#Fake images incorrectly classified: 39
+
+#KFold(n_splits=4, batch_size = 32, epochs = 5
+#train only last layer
+#Test Loss: 0.5492929220199585, Test Accuracy: 0.7992831468582153
+#Average accuracy: 88.79%
+#Classification Summary:
+#Real images correctly classified: 129
+#Real images incorrectly classified: 10
+#Fake images correctly classified: 94
+#Fake images incorrectly classified: 46
+#Classification Report:
+#              precision    recall  f1-score   support
+#
+#        Real       0.74      0.93      0.82       139
+#        Fake       0.90      0.67      0.77       140
+
+import random
+import numpy as np
+import os
+import pandas as pd
+import cv2
+import warnings
+from sklearn.model_selection import KFold
+import tensorflow as tf
+from sklearn.metrics import accuracy_score, confusion_matrix, ConfusionMatrixDisplay
+import matplotlib.pyplot as plt
+from sklearn.metrics import precision_score, recall_score, f1_score, classification_report
+
+# Ensure h5py is installed
+import h5py
+
+# Set the random seed for numpy, tensorflow, and python built-in random module
+np.random.seed(42)
+tf.random.set_seed(42)
+random.seed(42)
+
+warnings.filterwarnings("ignore", category=UserWarning, message=".*iCCP:.*")
+
+# Define data paths
+train_real_folder = 'datasets/training_set/real/'
+train_fake_folder = 'datasets/training_set/fake/'
+test_real_folder = 'datasets/test_set/real/'
+test_fake_folder = 'datasets/test_set/fake/'
+
+# Load train image paths and labels
+train_image_paths = []
+train_labels = []
+
+# Load train_real image paths and labels
+for filename in os.listdir(train_real_folder):
+    image_path = os.path.join(train_real_folder, filename)
+    label = 0  # Real images have label 0
+    train_image_paths.append(image_path)
+    train_labels.append(label)
+
+# Load train_fake image paths and labels
+for filename in os.listdir(train_fake_folder):
+    image_path = os.path.join(train_fake_folder, filename)
+    label = 1  # Fake images have label 1
+    train_image_paths.append(image_path)
+    train_labels.append(label)
+
+# Load test image paths and labels
+test_image_paths = []
+test_labels = []
+
+# Load test_real image paths and labels
+for filename in os.listdir(test_real_folder):
+    image_path = os.path.join(test_real_folder, filename)
+    label = 0  # Assuming test real images are all real (label 0)
+    test_image_paths.append(image_path)
+    test_labels.append(label)
+
+# Load test_fake image paths and labels
+for filename in os.listdir(test_fake_folder):
+    image_path = os.path.join(test_fake_folder, filename)
+    label = 1  # Assuming test fake images are all fake (label 1)
+    test_image_paths.append(image_path)
+    test_labels.append(label)
+
+# Create DataFrames
+train_dataset = pd.DataFrame({'image_path': train_image_paths, 'label': train_labels})
+test_dataset = pd.DataFrame({'image_path': test_image_paths, 'label': test_labels})
+
+# Function to preprocess images
+def preprocess_image(image_path):
+    """Loads, resizes, and normalizes an image."""
+    image = cv2.imread(image_path)
+    resized_image = cv2.resize(image, (224, 224))  # Target size defined here
+    normalized_image = resized_image.astype(np.float32) / 255.0
+    return normalized_image
+
+# Preprocess all images and convert labels to numpy arrays
+X = np.array([preprocess_image(path) for path in train_image_paths])
+Y = np.array(train_labels)
+
+# Define ResNet50 model
+resnet_model = tf.keras.applications.ResNet50(weights='imagenet', include_top=False, input_shape=(224, 224, 3))
+
+# Freeze all layers except the last one
+for layer in resnet_model.layers[:-1]:
+    layer.trainable = False
+
+# Modify final layer
+x = resnet_model.output
+x = tf.keras.layers.Flatten()(x)
+predictions = tf.keras.layers.Dense(1, activation='sigmoid')(x)  # Binary classification
+
+# Create new model with modified top
+new_model = tf.keras.models.Model(inputs=resnet_model.input, outputs=predictions)
+
+# Compile the new model
+new_model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
+
+# Set parameters for cross-validation
+kf = KFold(n_splits=4, shuffle=True, random_state=42)
+batch_size = 32
+epochs = 5
+weights_file = 'model_2.weights.h5'
+# Lists to store accuracy and loss for each fold
+accuracy_per_fold = []
+loss_per_fold = []
+
+# Perform K-Fold Cross-Validation
+for train_index, val_index in kf.split(X):
+    X_train, X_val = X[train_index], X[val_index]
+    Y_train, Y_val = Y[train_index], Y[val_index]
+    
+    # Load weights if they exist
+    if os.path.exists(weights_file):
+        new_model.load_weights(weights_file)
+        print(f"Loaded weights from {weights_file}")
+    
+    # Train only the last layer
+    history = new_model.fit(X_train, Y_train, epochs=epochs, batch_size=batch_size, verbose=1, validation_data=(X_val, Y_val))
+    
+    # Save weights after training
+    new_model.save_weights(weights_file)
+    print(f"Saved weights to {weights_file}")
+    
+    # Evaluate the model on the validation data
+    val_loss, val_accuracy = new_model.evaluate(X_val, Y_val)
+    
+    # Store the accuracy score for this fold
+    accuracy_per_fold.append(val_accuracy)
+    loss_per_fold.append(val_loss)
+    print(f'Fold accuracy: {val_accuracy*100:.2f}%')
+    print(f'Fold loss: {val_loss:.4f}')
+    
+# Print average accuracy and loss across all folds
+print(f'\nAverage accuracy across all folds: {np.mean(accuracy_per_fold)*100:.2f}%')
+print(f'Average loss across all folds: {np.mean(loss_per_fold):.4f}')
+
+# Evaluate the preprocessed test images using the final model
+test_loss, test_accuracy = new_model.evaluate(np.array([preprocess_image(path) for path in test_image_paths]), np.array(test_labels))
+print(f"\nTest Loss: {test_loss}, Test Accuracy: {test_accuracy}")
+
+# Predict labels for the test set
+predictions = new_model.predict(np.array([preprocess_image(path) for path in test_image_paths]))
+predicted_labels = (predictions > 0.5).astype(int).flatten()
+
+# Summarize the classification results
+true_real_correct = np.sum((np.array(test_labels) == 0) & (predicted_labels == 0))
+true_real_incorrect = np.sum((np.array(test_labels) == 0) & (predicted_labels == 1))
+true_fake_correct = np.sum((np.array(test_labels) == 1) & (predicted_labels == 1))
+true_fake_incorrect = np.sum((np.array(test_labels) == 1) & (predicted_labels == 0))
+
+print("\nClassification Summary:")
+print(f"Real images correctly classified: {true_real_correct}")
+print(f"Real images incorrectly classified: {true_real_incorrect}")
+print(f"Fake images correctly classified: {true_fake_correct}")
+print(f"Fake images incorrectly classified: {true_fake_incorrect}")
+
+# Print detailed classification report
+print("\nClassification Report:")
+print(classification_report(test_labels, predicted_labels, target_names=['Real', 'Fake']))
+
+# Plot confusion matrix
+cm = confusion_matrix(test_labels, predicted_labels)
+disp = ConfusionMatrixDisplay(confusion_matrix=cm, display_labels=['Real', 'Fake'])
+disp.plot(cmap=plt.cm.Blues)
+plt.title("Confusion Matrix")
+plt.show()
+
+# Plot training & validation accuracy values
+plt.figure(figsize=(12, 4))
+plt.subplot(1, 2, 1)
+plt.plot(history.history['accuracy'])
+plt.plot(history.history['val_accuracy'])
+plt.title('Model accuracy')
+plt.ylabel('Accuracy')
+plt.xlabel('Epoch')
+plt.legend(['Train', 'Validation'], loc='upper left')
+plt.xticks(np.arange(0, len(history.history['accuracy']), step=1), np.arange(1, len(history.history['accuracy']) + 1, step=1))
+
+# Plot training & validation loss values
+plt.subplot(1, 2, 2)
+plt.plot(history.history['loss'])
+plt.plot(history.history['val_loss'])
+plt.title('Model loss')
+plt.ylabel('Loss')
+plt.xlabel('Epoch')
+plt.legend(['Train', 'Validation'], loc='upper left')
+plt.xticks(np.arange(0, len(history.history['loss']), step=1), np.arange(1, len(history.history['loss']) + 1, step=1))
+
+plt.tight_layout()
+plt.show()
+
+# Plot validation accuracy and loss per fold
+plt.figure(figsize=(12, 4))
+plt.subplot(1, 2, 1)
+plt.plot(range(1, kf.get_n_splits() + 1), accuracy_per_fold, marker='o')
+plt.title('Validation Accuracy per Fold')
+plt.xlabel('Fold')
+plt.ylabel('Accuracy')
+
+plt.subplot(1, 2, 2)
+plt.plot(range(1, kf.get_n_splits() + 1), loss_per_fold, marker='o')
+plt.title('Validation Loss per Fold')
+plt.xlabel('Fold')
+plt.ylabel('Loss')
+
+plt.tight_layout()
+plt.show()

model_2.weights.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb30a1a6a3824eb47bc067d5c9b17bb8944c4efe2108a81973fb4aeee52d817
+size 96013592