app.py

import streamlit as st
from cvzone.PoseModule import PoseDetector
import cv2
import time
from gtts import gTTS
import os

# Initialize the PoseDetector class
detector = PoseDetector(staticMode=False,
                        modelComplexity=1,
                        smoothLandmarks=True,
                        enableSegmentation=True,
                        smoothSegmentation=True,
                        detectionCon=0.7,
                        trackCon=0.4)

# Global variables for position checking
hold_time = 0
start_time = None
cooldown_time = 2
last_spoken_time = time.time() - cooldown_time

# Define text labels
correct_position_text = "Correct Knee Hug Position"
incorrect_position_text = "Incorrect Position"
hold_complete_text = "Hold Complete!"
font_scale = 0.6
correct_color = (0, 255, 0)  # Green for correct
incorrect_color = (0, 0, 255)  # Red for incorrect

# Function for TTS using gTTS
def speak(text, filename="output.mp3"):
    tts = gTTS(text)
    tts.save(filename)
    st.audio(filename, format="audio/mp3")

# Function to detect the pose
def run_pose_detection():
    global hold_time, start_time, last_spoken_time

    # Capture webcam
    cap = cv2.VideoCapture(0)
    
    while cap.isOpened():
        success, img = cap.read()
        if not success:
            break

        # Process the frame
        img = detector.findPose(img)
        lmList, bboxInfo = detector.findPosition(img, draw=False, bboxWithHands=False)

        if lmList:
            # Calculate angles for the knees
            LHK_angle, img = detector.findAngle(lmList[23][0:2], lmList[25][0:2], lmList[27][0:2], img=img, color=(255, 0, 0), scale=1)
            RHK_angle, img = detector.findAngle(lmList[24][0:2], lmList[26][0:2], lmList[28][0:2], img=img, color=(255, 0, 0), scale=1)

            # Define the correct position range
            if 170 <= LHK_angle <= 340 and 300 <= RHK_angle <= 340:
                cv2.putText(img, correct_position_text, (50, 50), cv2.FONT_HERSHEY_SIMPLEX, font_scale, correct_color, 2, cv2.LINE_AA)
                
                if start_time is None:
                    start_time = time.time()
                else:
                    hold_time = time.time() - start_time

                countdown_text = f"Hold for: {int(5 - hold_time)} seconds"
                cv2.putText(img, countdown_text, (50, 200), cv2.FONT_HERSHEY_SIMPLEX, font_scale, correct_color, 2, cv2.LINE_AA)

                current_time = time.time()
                if current_time - last_spoken_time >= cooldown_time:
                    speak("Correct position!")
                    last_spoken_time = current_time

                if hold_time >= 5:
                    cv2.putText(img, hold_complete_text, (50, 250), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0, 255, 255), 2, cv2.LINE_AA)
                    if current_time - last_spoken_time >= cooldown_time:
                        speak("Hold Complete!")
                        last_spoken_time = current_time
            else:
                cv2.putText(img, incorrect_position_text, (50, 50), cv2.FONT_HERSHEY_SIMPLEX, font_scale, incorrect_color, 2, cv2.LINE_AA)
                hold_time = 0
                start_time = None

        # Convert to RGB for Streamlit
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

        # Show frame in Streamlit
        st.image(img, channels="RGB")

        # Streamlit refresh every 30ms
        time.sleep(0.03)

# Streamlit user interface
st.title("Physaio - Knee Hug Pose Detection")
st.write("This app detects the correctness of your knee hug pose using your webcam.")

# Start button for pose detection
if st.button("Start Pose Detection"):
    run_pose_detection()

# Display footer
st.write("Ensure your webcam is connected and working properly for this application.")