Gesture volume control with python opencv. Upgraded version of Murtaza's workshop

HandControl.py

import cv2
import mediapipe as mp
import numpy as np
import math
from ctypes import cast, POINTER
from comtypes import CLSCTX_ALL
from pycaw.pycaw import AudioUtilities, IAudioEndpointVolume

np.set_printoptions(precision=3)

# short names to reference
mpDrawing = mp.solutions.drawing_utils
mpHands = mp.solutions.hands
LMS = mpHands.HandLandmark

# pycaw base code for master volume control
devices = AudioUtilities.GetSpeakers()
interface = devices.Activate(IAudioEndpointVolume._iid_, CLSCTX_ALL, None)
volume = cast(interface, POINTER(IAudioEndpointVolume))

# take a video from camera with id = 0
capture = cv2.VideoCapture(0)
hands = mpHands.Hands(min_detection_confidence=0.7, max_num_hands=1)

while capture.isOpened():
    success, image = capture.read()

    # ignoring empty frame
    if not success:
        continue
    
    # convert image to RGB format
    imageRGB = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
    processedHands = hands.process(imageRGB)
    
    # check if hands were found
    if processedHands.multi_hand_landmarks is not None:
        # loop through landmarks for each hand
        for handLandmarks in processedHands.multi_hand_landmarks:
            mp.solutions.drawing_utils.draw_landmarks(image, handLandmarks, mpHands.HAND_CONNECTIONS)

            # Get coordinates for thumb tip and index finger tip. Also get coordinates for wrist to use in the Pythagoras rule
            # Assuming that at maximum extension the fingers form a triangle we can calculate max distance between tips
            # Using that fact we can normalize distance between finger tips
            lm = handLandmarks.landmark
            thumb = np.array([lm[LMS.THUMB_TIP].x,        lm[LMS.THUMB_TIP].y])
            index = np.array([lm[LMS.INDEX_FINGER_TIP].x, lm[LMS.INDEX_FINGER_TIP].y])
            wrist = np.array([lm[LMS.WRIST].x,            lm[LMS.WRIST].y])

            # Get rough length of both fingers
            indexLength = np.linalg.norm(index - wrist)
            thumbLength = np.linalg.norm(thumb - wrist)

            # Now calculate max distance (also lower it down because it rough calculations)
            maxDistance = 0.7 * math.hypot(indexLength, thumbLength)

            # Calculate current distance
            length = np.linalg.norm(index - thumb)

            # Map length value to value from -VOLUME_SPEED to VOLUME_SPEED
            VOLUME_SPEED = 0.1
            diff = np.interp(length, (0.0, maxDistance), [-VOLUME_SPEED, VOLUME_SPEED])

            # Get volume in scalar (not in decibels) and set it in scalars
            currentVolume = volume.GetMasterVolumeLevelScalar()
            newVolume = min(max((currentVolume + diff), 0.0), 1.0)
            volume.SetMasterVolumeLevelScalar(newVolume, None)

    cv2.imshow('Image', image)

    if cv2.waitKey(5) & 0xFF == 27:
        break
        
        
# Finalization
capture.release()
cv2.destroyAllWindows()