"""
Realtime Detection untuk Kata dengan visualisasi MediaPipe landmarks.

Proses sesuai README:
1. Webcam
2. MediaPipe
3. Landmark
4. Normalisasi
5. Window 30 Frame
6. Model TCN
7. Prediksi
8. Tampilkan hasil
"""

import sys
import cv2
from pathlib import Path

sys.path.insert(0, str(Path(__file__).parent.parent))

from utils.inference import GestureRecognizer
from utils.config import MODEL_DIR, WINDOW_SIZE


def run_realtime(category: str = "kata", camera_id: int = 0):
    """
    Run real-time recognition dengan visualisasi landmarks.

    Args:
        category: Nama kategori
        camera_id: Camera index
    """
    model_path = MODEL_DIR / "kata" / "model_kata.pt"
    encoder_path = MODEL_DIR / "kata" / "label_encoder.json"

    if not model_path.exists():
        print(f"Error: Model tidak ditemukan di {model_path}")
        print("Jalankan training terlebih dahulu!")
        return

    # Initialize recognizer
    recognizer = GestureRecognizer(
        model_path,
        encoder_path,
        window_size=WINDOW_SIZE,
        draw_landmarks=True
    )

    # Open camera
    cap = cv2.VideoCapture(camera_id)

    if not cap.isOpened():
        print(f"Error: Tidak dapat membuka camera {camera_id}")
        return

    print("\n" + "="*60)
    print(f"REALTIME RECOGNITION - {category.upper()}")
    print("="*60)
    print("Tekan 'q' untuk quit")
    print("Tekan 'r' untuk reset buffer")
    print("Tekan 'l' untuk toggle landmark visibility")
    print("="*60 + "\n")

    show_landmarks = True

    try:
        while True:
            ret, frame = cap.read()

            if not ret:
                print("Error: Tidak dapat membaca frame")
                break

            # Draw landmarks
            if show_landmarks:
                frame = recognizer.draw_landmarks_on_frame(frame)

            # Process frame
            label, confidence = recognizer.process_frame(frame)

            # Draw prediction
            if label is not None:
                text = f"{label} ({confidence*100:.1f}%)"
                cv2.putText(frame, text, (10, 50),
                           cv2.FONT_HERSHEY_SIMPLEX, 1.5, (0, 255, 0), 3)

            # Draw buffer status
            buffer_len = len(recognizer.buffer)
            status = f"Buffer: {buffer_len}/{recognizer.window_size}"
            cv2.putText(frame, status, (10, 100),
                       cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 255, 255), 2)

            # Draw landmark status
            landmark_status = "Landmarks: ON" if show_landmarks else "Landmarks: OFF"
            cv2.putText(frame, landmark_status, (10, 130),
                       cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 255, 255), 2)

            # Draw detection status
            if label is None:
                cv2.putText(frame, "Tunjukkan gestur isyarat...", (10, 160),
                           cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)

            # Show frame
            cv2.imshow(f"Realtime Recognition - {category}", frame)

            # Check key
            key = cv2.waitKey(1) & 0xFF
            if key == ord('q'):
                break
            elif key == ord('r'):
                recognizer.reset()
                print("Buffer reset")
            elif key == ord('l'):
                show_landmarks = not show_landmarks
                print(f"Landmarks: {'ON' if show_landmarks else 'OFF'}")

    finally:
        cap.release()
        cv2.destroyAllWindows()
        recognizer.close()


def main():
    import argparse
    parser = argparse.ArgumentParser(description="Realtime recognition kata")
    parser.add_argument("--camera", type=int, default=0, help="Camera index")
    args = parser.parse_args()
    run_realtime("kata", args.camera)


if __name__ == "__main__":
    main()