display image fractal like range(5)

import numpy as np
import matplotlib.pyplot as plt
import tensorflow as tf
from tqdm.notebook import tqdm_notebook
from IPython.display import display, Javascript
from google.colab import files
import os
import shutil
import ast
from sklearn.metrics import confusion_matrix, accuracy_score, precision_score, recall_score, f1_score
import seaborn as sns
from skimage.transform import resize
import sys

display(Javascript('IPython.OutputArea.auto_scroll_threshold = 9999;'))

label_colors = {0: [0, 128, 0], 1: [255, 0, 0]}
label_colors_testing = {0: [0, 128, 0], 1: [255, 0, 0]}

%matplotlib inline

def create_image(data, predictions, label_colors, column_min_vals, column_max_vals):
    num_rows, num_columns = len(data), len(data[0])
    image = np.zeros((num_rows, num_columns + 1, 3), dtype=np.uint8)
    for i in range(num_rows):
        for j in range(num_columns):
            pixel_value = int(np.interp(data[i][j], [column_min_vals[j], column_max_vals[j]], [0, 255]))
            image[i, j] = np.array([pixel_value] * 3)
        image[i, -1] = label_colors[predictions[i]]
    return image

def create_imageN(data, predictions, label_colors, column_min_vals, column_max_vals):
    num_training_rows = len(data)
    num_columns = len(data[0])
    image_training = np.zeros((num_training_rows, num_columns + 1, 3), dtype=np.uint8)
    for i in range(num_training_rows):
        for j in range(num_columns):
            pixel_value = int(np.interp(data[i][j], [column_min_vals[j], column_max_vals[j]], [0, 255]))
            image_training[i, j] = np.array([pixel_value] * 3)
        image_training[i, -1] = label_colors[int(predictions[i])]
    return image_training

def create_cnn_model(input_shape):
    model = tf.keras.Sequential([
        tf.keras.layers.InputLayer(input_shape=input_shape),
        tf.keras.layers.Conv2D(filters=32, kernel_size=(3, 3), activation='relu'),
        tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),
        tf.keras.layers.Dropout(0.25),
        tf.keras.layers.Flatten(),
        tf.keras.layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)),
        tf.keras.layers.Dropout(0.5),
        tf.keras.layers.Dense(1, activation='sigmoid')
    ])
    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

new_persons_results = [
    [0.030391238492519845, 0.23021081913032299, 0.4743575198860915, 0.639395348276238],
    [0.19790381537769108, 0.37639843860181527, 0.5676528538456297, 0.716530820399044],
    [0.0035245462826666075, 0.23127629815305784, 0.4802171123709532, 0.6591272725083992],
    [0.059230621364548486, 0.24424510845680134, 0.442553808602372, 0.6891856336835676],
    [0.05536813173866345, 0.2538888869331579, 0.47861285542743165, 0.6200559751500355],
    [0.1300359168058454, 0.38443677757577344, 0.5957238735056223, 0.795823160451845],
    [0.1743368240338569, 0.3713129035302336, 0.5640350202165867, 0.7213527928848786],
    [0.09173335232875372, 0.2559096689549753, 0.49527436563146954, 0.6970388573439903],
    [0.015235204378572087, 0.2284904031445293, 0.46613902406934005, 0.6917336579549159],
    [0.0011416656054787145, 0.24567669307188245, 0.4388400949432476, 0.667323193441009],
    [0.11776711, 0.17521301, 0.5074825,  0.8509191 ],
    [0.12314088, 0.27565651, 0.52214202, 0.77386896],
]

uploaded = files.upload()
for filename in uploaded.keys():
    original_path = f"/content/{filename}"
    destination_path = os.path.join("/content/", "/content/DATA2")
    shutil.move(original_path, destination_path)
    print(f"Soubor {filename} byl přesunut do {destination_path}")

file_path = '/content/DATA2'
with open(file_path, 'r') as file:
    code = file.read()

A_list = ast.literal_eval(code)
A = np.array(A_list)

labels = [results[-1] for results in A]
data = [results[:-1] for results in A]

num_training_rows = 50
num_testing_rows = 50
X_train, X_test, y_train, y_test = data[:num_training_rows], data[num_training_rows:num_training_rows+num_testing_rows], labels[:num_training_rows], labels[num_training_rows:num_training_rows+num_testing_rows]
X_train, X_test, y_train, y_test = data[:num_training_rows], data[:num_testing_rows], labels[:num_training_rows], labels[:num_testing_rows]

mean_values = np.mean(X_train, axis=0)
std_values = np.std(X_train, axis=0)
X_train_normalized = (X_train - mean_values) / std_values
X_test_normalized = (X_test - mean_values) / std_values

column_min_vals = np.min(X_train_normalized, axis=0)
column_max_vals = np.max(X_train_normalized, axis=0)

# Verify normalization
print("Mean of X_train_normalized (should be close to 0):", np.mean(X_train_normalized, axis=0))
print("Std of X_train_normalized (should be close to 1):", np.std(X_train_normalized, axis=0))

dnn_model = tf.keras.Sequential([
    tf.keras.layers.Dense(128, activation='relu', input_shape=(len(X_train[0]),)),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.2),  # Snížení dropout rate
    tf.keras.layers.Dense(64, activation='relu'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.2),
    tf.keras.layers.Dense(32, activation='relu'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.2),
    tf.keras.layers.Dense(16, activation='relu'),
    tf.keras.layers.BatchNormalization(),
    tf.keras.layers.Dropout(0.2),
    tf.keras.layers.Dense(1, activation='sigmoid')
])
dnn_model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001), loss='binary_crossentropy', metrics=['accuracy'])

# Generování dummy předpovědí pro tréninková data
train_predictions = [0] * len(X_train_normalized)  # Dummy předpovědi pro tréninkové obrázky

# Generování skutečných předpovědí pro testovací data
test_predictions = (dnn_model.predict(X_test_normalized) > 0.5).astype(int).flatten()

# Generování obrázků z normalizovaných dat
image_training = create_imageN(X_train_normalized, y_train, label_colors, column_min_vals, column_max_vals)
image_testing = create_image(X_test_normalized, test_predictions, label_colors_testing, column_min_vals, column_max_vals)

# Ověření počtu vygenerovaných obrázků
print("Počet tréninkových obrázků:", len(image_training))
print("Počet testovacích obrázků:", len(image_testing))
assert len(image_training) == 50, "Počet tréninkových obrázků by měl být 50."
assert len(image_testing) == len(X_test_normalized), "Počet testovacích obrázků by měl odpovídat počtu testovacích vzorků."

# Změna velikosti obrázků na pevnou velikost pro vstup do CNN
image_training_resized = [resize(img[:, :-1], (100, 100, 3)) for img in image_training]
image_testing_resized = [resize(img[:, :-1], (100, 100, 3)) for img in image_testing]

# Ověření změny velikosti obrázků
print("Tvar prvního tréninkového obrázku po změně velikosti:", image_training_resized[0].shape)
print("Tvar prvního testovacího obrázku po změně velikosti:", image_testing_resized[0].shape)

# Převod obrázků do formátu numpy pole
X_train_cnn = np.array(image_training_resized)
X_test_cnn = np.array(image_testing_resized)

# Ověření tvaru polí
print("Tvar X_train_cnn:", X_train_cnn.shape)
print("Tvar X_test_cnn:", X_test_cnn.shape)
assert X_train_cnn.shape == (50, 100, 100, 3), "Tvar X_train_cnn by měl být (50, 100, 100, 3)."

# Funkce pro zobrazení obrázku
def display_image(image, title="Image"):
    plt.imshow(image)
    plt.title(title)
    plt.axis('off')
    plt.show()

# Zobrazení několika tréninkových obrázků
for i in range(5):
    display_image(X_train_cnn[i], title=f"Tréninkový obrázek {i+1}")

# Trénování modelu DNN
dnn_accuracy_history = []
epochs = 500  # Reduced to 500 for quicker convergence

for epoch in tqdm_notebook(range(epochs)):
    history_dnn = dnn_model.fit(X_train_normalized, np.array(y_train), epochs=1, verbose=0, shuffle=False)
    dnn_accuracy_history.append(history_dnn.history['accuracy'][0])

    if epoch == 1:
        y_pred_after_2nd_epoch_dnn = dnn_model.predict(X_test_normalized)
        y_pred_binary_after_2nd_epoch_dnn = [1 if pred >= 0.5 else 0 for pred in y_pred_after_2nd_epoch_dnn]
        image_testing_before_2nd_epoch_dnn = create_image(X_test_normalized, y_pred_binary_after_2nd_epoch_dnn, label_colors_testing, column_min_vals, column_max_vals)

    if epoch >= epochs-1:
        print(f"HERE HERE Epoch: {epoch}, Epochs: {epochs}\n")
        sys.stdout.flush()

        # Iterate through new persons
        for idx, personNEW_results in enumerate(new_persons_results, start=1):
            assert len(personNEW_results) == len(X_train[0]), "Mismatch in the number of features."
            personNEW_results_normalized = (np.array(personNEW_results) - mean_values) / std_values
            personNEW_prediction_dnn = dnn_model.predict(np.array([personNEW_results_normalized]))
            personNEW_label_dnn = 1 if personNEW_prediction_dnn >= 0.5 else 0
            y_pred_after_50_epochs_dnn = dnn_model.predict(X_test_normalized)
            y_pred_binary_after_50_epochs_dnn = [1 if pred >= 0.5 else 0 for pred in y_pred_after_50_epochs_dnn]
            image_testing_after_50_epochs_dnn = create_image(X_test_normalized, y_pred_binary_after_50_epochs_dnn, label_colors_testing, column_min_vals, column_max_vals)
            image_personNEW_dnn = create_imageN([personNEW_results_normalized], [personNEW_label_dnn], label_colors, column_min_vals, column_max_vals)
            plt.figure(figsize=(5, 5))
            plt.imshow(image_personNEW_dnn)
            plt.title(f"New Person {idx} - DNN\nLabel: {personNEW_label_dnn}, Prediction: {personNEW_prediction_dnn}")
            plt.axis("off")
            plt.show()

# CNN Model
cnn_model = tf.keras.Sequential([
    tf.keras.layers.InputLayer(input_shape=(100, 100, 3)),
    tf.keras.layers.Conv2D(filters=32, kernel_size=(3, 3), activation='relu'),
    tf.keras.layers.MaxPooling2D(pool_size=(2, 2)),
    tf.keras.layers.Dropout(0.25),
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(64, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(1, activation='sigmoid')
])
cnn_model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

# Training CNN Model
cnn_accuracy_history = []

for epoch in tqdm_notebook(range(epochs)):
    history_cnn = cnn_model.fit(X_train_cnn, np.array(y_train), epochs=1, verbose=0, shuffle=False)
    cnn_accuracy_history.append(history_cnn.history['accuracy'][0])

    if epoch == 1:
        y_pred_after_2nd_epoch_cnn = cnn_model.predict(X_test_cnn)
        y_pred_binary_after_2nd_epoch_cnn = [1 if pred >= 0.5 else 0 for pred in y_pred_after_2nd_epoch_cnn]
        image_testing_before_2nd_epoch_cnn = create_image(X_test_normalized, y_pred_binary_after_2nd_epoch_cnn, label_colors_testing, column_min_vals, column_max_vals)

    if epoch >= epochs-1:
        print(f"HERE HERE Epoch: {epoch}, Epochs: {epochs}\n")
        sys.stdout.flush()

        # Iterate through new persons
        for idx, personNEW_results in enumerate(new_persons_results, start=1):
            assert len(personNEW_results) == len(X_train[0]), "Mismatch in the number of features."
            personNEW_results_normalized = (np.array(personNEW_results) - mean_values) / std_values
            image_personNEW = create_imageN([personNEW_results_normalized], [0], label_colors, column_min_vals, column_max_vals)
            image_personNEW_resized = resize(image_personNEW[:, :-1], (100, 100, 3))
            personNEW_prediction_cnn = cnn_model.predict(np.array([image_personNEW_resized]))
            personNEW_label_cnn = 1 if personNEW_prediction_cnn >= 0.5 else 0
            y_pred_after_50_epochs_cnn = cnn_model.predict(X_test_cnn)
            y_pred_binary_after_50_epochs_cnn = [1 if pred >= 0.5 else 0 for pred in y_pred_after_50_epochs_cnn]
            image_testing_after_50_epochs_cnn = create_image(X_test_normalized, y_pred_binary_after_50_epochs_cnn, label_colors_testing, column_min_vals, column_max_vals)
            image_personNEW_cnn = create_imageN([personNEW_results_normalized], [personNEW_label_cnn], label_colors, column_min_vals, column_max_vals)
            plt.figure(figsize=(5, 5))
            plt.imshow(image_personNEW_cnn)
            plt.title(f"New Person {idx} - CNN\nLabel: {personNEW_label_cnn}, Prediction: {personNEW_prediction_cnn}")
            plt.axis("off")
            plt.show()

# Display the images
plt.figure(figsize=(25, 15))
plt.subplot(2, 2, 1)
plt.imshow(image_training)
plt.title("Training Data")
plt.axis("off")

plt.subplot(2, 2, 2)
plt.imshow(image_testing_before_2nd_epoch_dnn)
plt.title("Testing Data (2nd Epoch) - DNN")
plt.axis("off")

plt.subplot(2, 2, 3)
plt.imshow(image_testing_after_50_epochs_dnn)
plt.title(f"Testing Data ({epochs} Epochs) - DNN")
plt.axis("off")

plt.subplot(2, 2, 4)
plt.imshow(image_personNEW_dnn)
plt.title(f"New Person - DNN\nLabel: {personNEW_label_dnn},[{personNEW_prediction_dnn}]")
plt.axis("off")

plt.figure(figsize=(12, 5))
plt.plot(range(1, epochs + 1), dnn_accuracy_history, marker='o')
plt.title('DNN Accuracy Over Epochs')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.grid()

plt.figure(figsize=(25, 15))
plt.subplot(2, 2, 1)
plt.imshow(image_training)
plt.title("Training Data")
plt.axis("off")

plt.subplot(2, 2, 2)
plt.imshow(image_testing_before_2nd_epoch_cnn)
plt.title("Testing Data (2nd Epoch) - CNN")
plt.axis("off")

plt.subplot(2, 2, 3)
plt.imshow(image_testing_after_50_epochs_cnn)
plt.title(f"Testing Data ({epochs} Epochs) - CNN")
plt.axis("off")

plt.subplot(2, 2, 4)
plt.imshow(image_personNEW_cnn)
plt.title(f"New Person - CNN\nLabel: {personNEW_label_cnn},[{personNEW_prediction_cnn}]")
plt.axis("off")

plt.figure(figsize=(12, 5))
plt.plot(range(1, epochs + 1), cnn_accuracy_history, marker='o')
plt.title('CNN Accuracy Over Epochs')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.grid()

# Confusion Matrix and Performance Metrics for DNN
dnn_predictions = (dnn_model.predict(X_test_normalized) > 0.5).astype(int)
dnn_conf_matrix = confusion_matrix(y_test, dnn_predictions)
print(f"Confusion Matrix for DNN:\n{dnn_conf_matrix}")
dnn_accuracy = accuracy_score(y_test, dnn_predictions)
dnn_precision = precision_score(y_test, dnn_predictions)
dnn_recall = recall_score(y_test, dnn_predictions)
dnn_f1 = f1_score(y_test, dnn_predictions)
print(f"DNN Accuracy: {dnn_accuracy:.4f}")
print(f"DNN Precision: {dnn_precision:.4f}")
print(f"DNN Recall: {dnn_recall:.4f}")
print(f"DNN F1 Score: {dnn_f1:.4f}")

# Confusion Matrix and Performance Metrics for CNN
cnn_predictions = (cnn_model.predict(X_test_cnn) > 0.5).astype(int)
print("PP:",X_test_cnn)
cnn_conf_matrix = confusion_matrix(y_test, cnn_predictions)
print(f"Confusion Matrix for CNN:\n{cnn_conf_matrix}")
cnn_accuracy = accuracy_score(y_test, cnn_predictions)
cnn_precision = precision_score(y_test, cnn_predictions)
cnn_recall = recall_score(y_test, cnn_predictions)
cnn_f1 = f1_score(y_test, cnn_predictions)
print(f"CNN Accuracy: {cnn_accuracy:.4f}")
print(f"CNN Precision: {cnn_precision:.4f}")
print(f"CNN Recall: {cnn_recall:.4f}")
print(f"CNN F1 Score: {cnn_f1:.4f}")

# Display confusion matrices
plt.figure(figsize=(12, 5))

plt.subplot(1, 2, 1)
sns.heatmap(dnn_conf_matrix, annot=True, fmt='d', cmap='Blues')
plt.xlabel('Predicted')
plt.ylabel('Actual')
plt.title('DNN Confusion Matrix')

plt.subplot(1, 2, 2)
sns.heatmap(cnn_conf_matrix, annot=True, fmt='d', cmap='Blues')
plt.xlabel('Predicted')
plt.ylabel('Actual')
plt.title('CNN Confusion Matrix')

plt.show()

# Optimalizace nového vektoru pro predikci co nejblíže 0.52
target_prediction = 0.52
input_shape = 4
new_vector = np.random.randn(input_shape)
new_vector = tf.Variable(new_vector, dtype=tf.float32)

optimizer = tf.optimizers.Adam(learning_rate=0.1)

def loss_function():
    prediction = dnn_model(tf.reshape(new_vector, (1, -1)))
    return tf.abs(prediction - target_prediction)

# Gradientní sestup
for _ in range(1000):
    optimizer.minimize(loss_function, [new_vector])

# Denormalizace výsledného vektoru
result_vector = new_vector.numpy()
denormalized_vector = result_vector * std_values + mean_values
result_prediction = dnn_model.predict(result_vector.reshape(1, -1))

print("Výsledný vektor (normalizovaný):", result_vector)
print("Výsledný vektor (denormalizovaný):", denormalized_vector)
print("Predikce výsledného vektoru:", result_prediction)