motion.py

from ultralytics import YOLO
import cv2
import time
import threading
import pygame

# Load YOLO models
model_yawn = YOLO(r"C:\Users\mihar\OneDrive\Documents\Deep Leaning Project\ultralytics\motion.pt")
model_eyes = YOLO(r"D:\Users\mihar\Downloads\best (7).pt")

# Audio control variables
close_alert_playing = False
audio_lock = threading.Lock()

def play_close_audio_once():
    global close_alert_playing
    with audio_lock:  # Lock to ensure one audio at a time
        pygame.mixer.init()
        pygame.mixer.music.load(r"C:\Users\mihar\OneDrive\Documents\Deep Leaning Project\ultralytics\voice1.mp3")
        pygame.mixer.music.play()
        while pygame.mixer.music.get_busy():
            time.sleep(0.1)
        close_alert_playing = False  # Reset only after audio is finished

def detect_with_model(model, frame, conf=0.25, label_color=(0, 255, 0)):
    detections = []
    results = model(frame, conf=conf, imgsz=640, device='cpu')  # imgsz=320 to improve speed

    for result in results:
        if result.boxes is not None:
            boxes = result.boxes.xyxy.cpu().numpy()
            scores = result.boxes.conf.cpu().numpy()
            class_ids = result.boxes.cls.cpu().numpy().astype(int)
            for i, box in enumerate(boxes):
                detections.append({
                    'box': box,
                    'score': scores[i],
                    'class_id': class_ids[i],
                    'class_name': model.names[class_ids[i]],
                    'color': label_color
                })
    return detections

def video_detectionn(frame, conf=0.25):
    global close_alert_playing
    start_time = time.time()
    output_frame = frame.copy()

    # Run both models
    detections_yawn = detect_with_model(model_yawn, frame, conf, (255, 0, 0))
    detections_eyes = detect_with_model(model_eyes, frame, conf, (0, 255, 0))

    detected_class_names = []
    total_detections = 0
    close_confident_count = 0  # Count 'close' with confidence > 0.50

    # Process detections
    for det in detections_yawn + detections_eyes:
        x1, y1, x2, y2 = map(int, det['box'])
        label = f"{det['class_name']} {det['score']:.2f}"
        cv2.rectangle(output_frame, (x1, y1), (x2, y2), det['color'], 2)
        cv2.putText(output_frame, label, (x1, y1 - 10),
                    cv2.FONT_HERSHEY_SIMPLEX, 0.6, det['color'], 2)
        detected_class_names.append(det['class_name'])
        total_detections += 1

        if det['class_name'].lower() == 'close' and det['score'] > 0.5:
            close_confident_count += 1

    # Trigger audio if two or more 'close' objects detected with score > 0.50
    if close_confident_count >= 2 and not close_alert_playing:
        close_alert_playing = True
        threading.Thread(target=play_close_audio_once).start()

    # FPS calculation
    fps_x = int(1 / (time.time() - start_time + 1e-5))
    cv2.putText(output_frame, f'Total Detected: {total_detections}', (10, 30),
                cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
    cv2.putText(output_frame, f'FPS: {fps_x}', (10, 60),
                cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)

    return [(output_frame, fps_x, output_frame.shape, total_detections)]