import numpy as np
import cv2
import os
import joblib
from skimage.feature import local_binary_pattern
from sklearn.neighbors import KNeighborsClassifier
from sklearn.metrics import accuracy_score

# Fungsi ekstraksi fitur LBP
def extract_lbp_features(image, radius=1, n_points=8, method='uniform'):
    lbp = local_binary_pattern(image, n_points, radius, method)
    (hist, _) = np.histogram(lbp.ravel(), bins=np.arange(0, n_points + 3), range=(0, n_points + 2))
    # Normalize the histogram
    hist = hist.astype("float")
    hist /= (hist.sum() + 1e-7)
    return hist

# Fungsi deteksi wajah dan pemotongan
def detect_and_crop_faces(image_path):
    # Baca gambar
    original_image = cv2.imread(image_path)
    gray_image = cv2.cvtColor(original_image, cv2.COLOR_BGR2GRAY)

    # Peningkatan kontras
    gray_image = cv2.equalizeHist(gray_image)

    # Muat pre-trained Haar Cascade untuk deteksi wajah
    face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')

    # Deteksi wajah pada gambar dengan parameter yang disesuaikan
    faces = face_cascade.detectMultiScale(gray_image, scaleFactor=1.2, minNeighbors=10, minSize=(50, 50))

    # Cek jumlah wajah yang terdeteksi
    print(f'{image_path}: {len(faces)} face(s) detected')

    # Loop melalui setiap wajah yang terdeteksi, potong, dan tampilkan
    for i, (x, y, w, h) in enumerate(faces, 1):
        # Pemotongan (Cropping) wajah
        cropped_face = original_image[y:y + h, x:x + w]

        # # Tampilkan gambar wajah yang terpotong
        # cv2.imshow(f'Cropped Face {i}', cropped_face)
        #
        # # Tunggu 1 detik sebelum menampilkan gambar berikutnya
        # cv2.waitKey(1000)

    return faces

# Persiapkan data latih dan label
X_train = []  # Menampung fitur-fitur ekstraksi
y_train = []  # Menampung label kelas

# Path root folder
root_folder = 'images/trainface/'

# Iterasi melalui setiap folder
for person_folder in os.listdir(root_folder):
    person_folder_path = os.path.join(root_folder, person_folder)

    # Pastikan path adalah direktori
    if os.path.isdir(person_folder_path):
        # Iterasi melalui setiap gambar dalam folder
        for image_name in os.listdir(person_folder_path):
            image_path = os.path.join(person_folder_path, image_name)

            # Pastikan path adalah file gambar
            if os.path.isfile(image_path):
                # Deteksi wajah dan cropping
                faces = detect_and_crop_faces(image_path)

                # Baca kembali gambar setelah deteksi wajah
                original_image = cv2.imread(image_path)

                # Loop melalui setiap wajah yang terdeteksi, potong, dan ekstraksi LBP
                for i, (x, y, w, h) in enumerate(faces, 1):
                    # Pemotongan (Cropping) wajah
                    cropped_face = original_image[y:y + h, x:x + w]

                    # Resize gambar menjadi 100x100 piksel
                    resized_face = cv2.resize(cropped_face, (100, 100))

                    # Konversi gambar resized ke grayscale
                    gray_resized_face = cv2.cvtColor(resized_face, cv2.COLOR_BGR2GRAY)

                    # Normalisasi intensitas
                    normalized_face = cv2.normalize(gray_resized_face, None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX)

                    # Ekstraksi fitur LBP
                    lbp_features = extract_lbp_features(normalized_face, radius=1, n_points=8, method='uniform')

                    # Tampilkan nilai ekstraksi LBP ke konsol
                    print(f'LBP Features for face {i}: {lbp_features}')

                    # Tambahkan fitur dan label ke data latih
                    X_train.append(lbp_features)
                    y_train.append(person_folder)  # Label kelas adalah nama folder

# Inisialisasi dan latih model K-NN
knn_classifier = KNeighborsClassifier(n_neighbors=1)  # Atur parameter K sesuai kebutuhan
knn_classifier.fit(X_train, y_train)

# Prediksi kelas pada data latih
y_train_pred = knn_classifier.predict(X_train)

# Hitung dan cetak akurasi pada data latih
train_accuracy = accuracy_score(y_train, y_train_pred)
print(f"Training Accuracy: {train_accuracy:.2f}")

# Simpan model ke file
joblib.dump(knn_classifier, 'facemodel.joblib')