import streamlit as st import cv2 from ultralytics import YOLO import cvzone import math import pygame # Initialize pygame mixer pygame.mixer.init() # Load sound alert_sound = pygame.mixer.Sound('siren-alert-96052.mp3') # Using raw string # Load the model model = YOLO('best.pt') # Reading the classes classnames = ['Drowsy', 'Awake'] # Streamlit UI st.title("Real-Time Drowsiness Detection") # Layout col1, col2 = st.columns(2) with col1: start_button = st.button('Start Webcam') with col2: stop_button = st.button('Stop Webcam') stframe = st.empty() status_text = st.empty() message_text = st.empty() if start_button: cap = cv2.VideoCapture(0) drowsy_count = 0 # Counter for consecutive "Drowsy" detections while cap.isOpened(): ret, frame = cap.read() if not ret: status_text.write("Failed to grab frame") break frame = cv2.resize(frame, (640, 480)) # Run the model on the frame result = model(frame, stream=True) # Flag to track if "Drowsy" is detected in this frame drowsy_detected = False # Getting bbox, confidence, and class name information to work with for info in result: boxes = info.boxes for box in boxes: confidence = box.conf[0] confidence = math.ceil(confidence * 100) Class = int(box.cls[0]) if confidence > 50: x1, y1, x2, y2 = box.xyxy[0] x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2) cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 5) cvzone.putTextRect(frame, f'{classnames[Class]} {confidence}%', [x1 + 8, y1 + 100], scale=1.5, thickness=2) if classnames[Class] == 'Drowsy': drowsy_detected = True # Increment the counter if "Drowsy" is detected, otherwise reset the counter if drowsy_detected: drowsy_count += 1 status_text.write("Drowsiness detected!") else: drowsy_count = 0 status_text.write("Monitoring...") # Play alert sound and send message if "Drowsy" is detected 3 or more times if drowsy_count >= 3: pygame.mixer.Sound.play(alert_sound) message_text.write("**Be careful!** Drowsiness detected multiple times!") drowsy_count = 0 # Reset the counter after playing the sound # Convert image back to RGB for Streamlit frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) # Display the image stframe.image(frame, channels="RGB") # Check if stop button is pressed if stop_button: break cap.release() status_text.write("Webcam stopped.") message_text.write("")