In this video lesson we show you how to create simple Python classes to parse the data from mediapipe hand Landmarks and pose estimation. With these two classes, it is very easy to parse the most important data generated by mediapipe.

import cv2
print(cv2.__version__)

class mpPose:
    import mediapipe as mp
    def __init__(self,still=False,upperBody=False, smoothData=True, tol1=.5, tol2=.5):
        self.myPose=self.mp.solutions.pose.Pose(still,upperBody,smoothData,tol1,tol2)
    def Marks(self,frame):
        frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
        results=self.myPose.process(frameRGB)
        poseLandmarks=[]
        if results.pose_landmarks:
            for lm in results.pose_landmarks.landmark:            poseLandmarks.append((int(lm.x*width),int(lm.y*height)))
        return poseLandmarks

class mpHands:
    import mediapipe as mp
    def __init__(self,maxHands=2,tol1=.5,tol2=.5):
        self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)
    def Marks(self,frame):
        myHands=[]
        frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
        results=self.hands.process(frameRGB)
        if results.multi_hand_landmarks != None:
            for handLandMarks in results.multi_hand_landmarks:
                myHand=[]
                for landMark in handLandMarks.landmark:
                    myHand.append((int(landMark.x*width),int(landMark.y*height)))
                myHands.append(myHand)
        return myHands

width=1280
height=720
cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)
cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)
cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)
cam.set(cv2.CAP_PROP_FPS, 30)
cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))

findHands=mpHands(2)
findPose=mpPose()
while True:
    ignore,  frame = cam.read()
    frame=cv2.resize(frame,(width,height))
    handData=findHands.Marks(frame)
    for hand in handData:
        for ind in [0,5,6,7,8]:
            cv2.circle(frame,hand[ind],25,(255,0,255),3)
    poseData=findPose.Marks(frame)
    if len(poseData)!=0:
        cv2.circle(frame,poseData[0],5,(0,255,0),3)

    cv2.imshow('my WEBcam', frame)
    cv2.moveWindow('my WEBcam',0,0)
    if cv2.waitKey(1) & 0xff ==ord('q'):
        break
cam.release()

import cv2

print(cv2.__version__)

class mpPose:

import mediapipe as mp

def __init__(self,still=False,upperBody=False, smoothData=True, tol1=.5, tol2=.5):

self.myPose=self.mp.solutions.pose.Pose(still,upperBody,smoothData,tol1,tol2)

def Marks(self,frame):

frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)

results=self.myPose.process(frameRGB)

poseLandmarks=[]

if results.pose_landmarks:

for lm in results.pose_landmarks.landmark: poseLandmarks.append((int(lm.x*width),int(lm.y*height)))

return poseLandmarks

class mpHands:

import mediapipe as mp

def __init__(self,maxHands=2,tol1=.5,tol2=.5):

self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)

def Marks(self,frame):

myHands=[]

frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)

results=self.hands.process(frameRGB)

if results.multi_hand_landmarks != None:

for handLandMarks in results.multi_hand_landmarks:

myHand=[]

for landMark in handLandMarks.landmark:

myHand.append((int(landMark.x*width),int(landMark.y*height)))

myHands.append(myHand)

return myHands

width=1280

height=720

cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)

cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)

cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)

cam.set(cv2.CAP_PROP_FPS, 30)

cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))

findHands=mpHands(2)

findPose=mpPose()

while True:

ignore, frame = cam.read()

frame=cv2.resize(frame,(width,height))

handData=findHands.Marks(frame)

for hand in handData:

for ind in [0,5,6,7,8]:

cv2.circle(frame,hand[ind],25,(255,0,255),3)

poseData=findPose.Marks(frame)

if len(poseData)!=0:

cv2.circle(frame,poseData[0],5,(0,255,0),3)

cv2.imshow('my WEBcam', frame)

cv2.moveWindow('my WEBcam',0,0)

if cv2.waitKey(1) & 0xff ==ord('q'):

break

cam.release()

Artificial Intelligence, OpenCV, Python

Distinguish Between Right and Left Hands in MediaPipe

November 16, 2021 admin

In this video lesson we explore how to parse the data set returned from mediapipe to understand whether a given hand is a right hand or a left hand. Mediapipe has a method which determines the handedness of a found hand. We will show how to alter our data parsing class from earlier lessons to include the handedness of the found hands. Enjoy!

import cv2
print(cv2.__version__)

class mpHands:
    import mediapipe as mp
    def __init__(self,maxHands=2,tol1=.5,tol2=.5):
        self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)
    def Marks(self,frame):
        myHands=[]
        handsType=[]
        frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
        results=self.hands.process(frameRGB)
        if results.multi_hand_landmarks != None:
            #print(results.multi_handedness)
            for hand in results.multi_handedness:
                #print(hand)
                #print(hand.classification)
                #print(hand.classification[0])
                handType=hand.classification[0].label
                handsType.append(handType)
            for handLandMarks in results.multi_hand_landmarks:
                myHand=[]
                for landMark in handLandMarks.landmark:
                    myHand.append((int(landMark.x*width),int(landMark.y*height)))
                myHands.append(myHand)
        return myHands,handsType

width=1280
height=720
cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)
cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)
cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)
cam.set(cv2.CAP_PROP_FPS, 30)
cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))
findHands=mpHands(2)
while True:
    ignore,  frame = cam.read()
    frame=cv2.resize(frame,(width,height))
    handData, handsType=findHands.Marks(frame)
    for hand,handType in zip(handData,handsType):
        if handType=='Right':
            handColor=(255,0,0)
        if handType=='Left':
            handColor=(0,0,255)
        for ind in [0,5,6,7,8]:
            cv2.circle(frame,hand[ind],15,handColor,5)
    cv2.imshow('my WEBcam', frame)
    cv2.moveWindow('my WEBcam',0,0)
    if cv2.waitKey(1) & 0xff ==ord('q'):
        break
cam.release()

import cv2

print(cv2.__version__)

class mpHands:

import mediapipe as mp

def __init__(self,maxHands=2,tol1=.5,tol2=.5):

self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)

def Marks(self,frame):

myHands=[]

handsType=[]

frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)

results=self.hands.process(frameRGB)

if results.multi_hand_landmarks != None:

#print(results.multi_handedness)

for hand in results.multi_handedness:

#print(hand)

#print(hand.classification)

#print(hand.classification[0])

handType=hand.classification[0].label

handsType.append(handType)

for handLandMarks in results.multi_hand_landmarks:

myHand=[]

for landMark in handLandMarks.landmark:

myHand.append((int(landMark.x*width),int(landMark.y*height)))

myHands.append(myHand)

return myHands,handsType

width=1280

height=720

cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)

cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)

cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)

cam.set(cv2.CAP_PROP_FPS, 30)

cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))

findHands=mpHands(2)

while True:

ignore, frame = cam.read()

frame=cv2.resize(frame,(width,height))

handData, handsType=findHands.Marks(frame)

for hand,handType in zip(handData,handsType):

if handType=='Right':

handColor=(255,0,0)

if handType=='Left':

handColor=(0,0,255)

for ind in [0,5,6,7,8]:

cv2.circle(frame,hand[ind],15,handColor,5)

cv2.imshow('my WEBcam', frame)

cv2.moveWindow('my WEBcam',0,0)

if cv2.waitKey(1) & 0xff ==ord('q'):

break

cam.release()

Artificial Intelligence, OpenCV, Python

AI for Everyone: Developing a Gesture Based Arcade Game Based on Mediapipe and Python

November 11, 2021 admin

In this video lesson we introduce the concept of using OpenCV and Mediapipe to create gesture based arcade games in Python. In this lesson we suggest creating a gesture based version of the classic Pong arcade game. We show enough to get started, and your homework is to complete the game.

import cv2
print(cv2.__version__)

class mpHands:
    import mediapipe as mp
    def __init__(self,maxHands=2,tol1=.5,tol2=.5):
        self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)
    def Marks(self,frame):
        myHands=[]
        frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
        results=self.hands.process(frameRGB)
        if results.multi_hand_landmarks != None:
            for handLandMarks in results.multi_hand_landmarks:
                myHand=[]
                for landMark in handLandMarks.landmark:
                    myHand.append((int(landMark.x*width),int(landMark.y*height)))
                myHands.append(myHand)
        return myHands

width=1280
height=720
cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)
cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)
cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)
cam.set(cv2.CAP_PROP_FPS, 30)
cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))
findHands=mpHands(2)
paddleWidth=125
paddleHeight=25
paddleColor=(0,255,0)
while True:
    ignore,  frame = cam.read()
    frame=cv2.resize(frame,(width,height))
    handData=findHands.Marks(frame)
    for hand in handData:
        cv2.rectangle(frame,(int(hand[8][0]-paddleWidth/2),0),(int(hand[8][0]+paddleWidth/2),paddleHeight),paddleColor,-1)
    cv2.imshow('my WEBcam', frame)
    cv2.moveWindow('my WEBcam',0,0)
    if cv2.waitKey(1) & 0xff ==ord('q'):
        break
cam.release()

import cv2

print(cv2.__version__)

class mpHands:

import mediapipe as mp

def __init__(self,maxHands=2,tol1=.5,tol2=.5):

self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)

def Marks(self,frame):

myHands=[]

frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)

results=self.hands.process(frameRGB)

if results.multi_hand_landmarks != None:

for handLandMarks in results.multi_hand_landmarks:

myHand=[]

for landMark in handLandMarks.landmark:

myHand.append((int(landMark.x*width),int(landMark.y*height)))

myHands.append(myHand)

return myHands

width=1280

height=720

cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)

cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)

cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)

cam.set(cv2.CAP_PROP_FPS, 30)

cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))

findHands=mpHands(2)

paddleWidth=125

paddleHeight=25

paddleColor=(0,255,0)

while True:

ignore, frame = cam.read()

frame=cv2.resize(frame,(width,height))

handData=findHands.Marks(frame)

for hand in handData:

cv2.rectangle(frame,(int(hand[8][0]-paddleWidth/2),0),(int(hand[8][0]+paddleWidth/2),paddleHeight),paddleColor,-1)

cv2.imshow('my WEBcam', frame)

cv2.moveWindow('my WEBcam',0,0)

if cv2.waitKey(1) & 0xff ==ord('q'):

break

cam.release()

Uncategorized

Python 3D Graphics Tutorial 18: 3D Clock Model with Proper Hand Motion

November 10, 2021 admin

In this lesson we create an Analog Clock Model with proper hand motion. The code developed in the lesson is shown below:

from vpython import *
import numpy as np
import time
clockR=1
clockT=clockR/10
majorTickL=clockR/7
majorTickT=2*np.pi*clockR/400
majorTickW=clockT*1.2

minorTickL=clockR/12
minorTickT=2*np.pi*clockR/600
minorTickW=clockT*1.2

minuteHandL=clockR-majorTickL
minuteHandT=minuteHandL/25
minuteHandOffset=clockT/2+minuteHandT

hourHandL=.75*minuteHandL
hourHandT=minuteHandT*1.25
hourHandOffset=clockT+hourHandT

secondHandL=clockR-majorTickL/2
secondHandT=minuteHandL/50
secondHandOffset=clockT*1.5+minuteHandT

hubRadius=clockT/2

hourAngle=np.pi/2
minuteAngle=np.pi/2
minInc=.0001
hourInc=minInc/12
secondAngle=np.pi/2
secondInc=minInc*60

for theta in np.linspace(0,2*np.pi,13):
    majorTick=box(axis=vector(clockR*np.cos(theta),clockR*np.sin(theta),0), color=color.black, length=majorTickL, width=majorTickW, height=majorTickT, pos=vector((clockR-majorTickL/2)*np.cos(theta),(clockR-majorTickL/2)*np.sin(theta),0))
for theta in np.linspace(0,2*np.pi,61):
    minorTick=box(axis=vector(clockR*np.cos(theta),clockR*np.sin(theta),0), color=color.black, length=minorTickL, width=minorTickW, height=minorTickT, pos=vector((clockR-minorTickL/2)*np.cos(theta),(clockR-minorTickL/2)*np.sin(theta),0))

clockFace=cylinder(axis=vector(0,0,1),color=vector(0,1,.8),length=clockT, radius=clockR,pos=vector(0,0,-clockT/2))
minuteHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=minuteHandT,length=minuteHandL, pos=vector(0,0,minuteHandOffset))
hourHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=hourHandT,length=hourHandL, pos=vector(0,0,hourHandOffset))
secondHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=secondHandT,length=secondHandL, pos=vector(0,0,secondHandOffset))
hub=cylinder(axis=vector(0,0,1),color=color.red, radius=hubRadius, length= 2*clockT)
while True:
    rate(5000)
    hour=time.localtime(time.time())[3]
    if hour>12:
        hour=hour-12
    minute=time.localtime(time.time())[4]
    second=time.localtime(time.time())[5]
    hourAngle=-((hour+minute/60)/12)*2*np.pi+np.pi/2
    minuteAngle=-((minute+second/60)/60)*2*np.pi+np.pi/2
    secondAngle=-(second/60)*2*np.pi+np.pi/2
    print(second)
    hourHand.axis=vector(hourHandL*np.cos(hourAngle),hourHandL*np.sin(hourAngle),0)
    minuteHand.axis=vector(minuteHandL*np.cos(minuteAngle),minuteHandL*np.sin(minuteAngle),0)
    secondHand.axis=vector(secondHandL*np.cos(secondAngle),secondHandL*np.sin(secondAngle),0)

from vpython import *

import numpy as np

import time

clockR=1

clockT=clockR/10

majorTickL=clockR/7

majorTickT=2*np.pi*clockR/400

majorTickW=clockT*1.2

minorTickL=clockR/12

minorTickT=2*np.pi*clockR/600

minorTickW=clockT*1.2

minuteHandL=clockR-majorTickL

minuteHandT=minuteHandL/25

minuteHandOffset=clockT/2+minuteHandT

hourHandL=.75*minuteHandL

hourHandT=minuteHandT*1.25

hourHandOffset=clockT+hourHandT

secondHandL=clockR-majorTickL/2

secondHandT=minuteHandL/50

secondHandOffset=clockT*1.5+minuteHandT

hubRadius=clockT/2

hourAngle=np.pi/2

minuteAngle=np.pi/2

minInc=.0001

hourInc=minInc/12

secondAngle=np.pi/2

secondInc=minInc*60

for theta in np.linspace(0,2*np.pi,13):

majorTick=box(axis=vector(clockR*np.cos(theta),clockR*np.sin(theta),0), color=color.black, length=majorTickL, width=majorTickW, height=majorTickT, pos=vector((clockR-majorTickL/2)*np.cos(theta),(clockR-majorTickL/2)*np.sin(theta),0))

for theta in np.linspace(0,2*np.pi,61):

minorTick=box(axis=vector(clockR*np.cos(theta),clockR*np.sin(theta),0), color=color.black, length=minorTickL, width=minorTickW, height=minorTickT, pos=vector((clockR-minorTickL/2)*np.cos(theta),(clockR-minorTickL/2)*np.sin(theta),0))

clockFace=cylinder(axis=vector(0,0,1),color=vector(0,1,.8),length=clockT, radius=clockR,pos=vector(0,0,-clockT/2))

minuteHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=minuteHandT,length=minuteHandL, pos=vector(0,0,minuteHandOffset))

hourHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=hourHandT,length=hourHandL, pos=vector(0,0,hourHandOffset))

secondHand=arrow(axis=vector(0,1,0),color=color.red,shaftwidth=secondHandT,length=secondHandL, pos=vector(0,0,secondHandOffset))

hub=cylinder(axis=vector(0,0,1),color=color.red, radius=hubRadius, length= 2*clockT)

while True:

rate(5000)

hour=time.localtime(time.time())[3]

if hour>12:

hour=hour-12

minute=time.localtime(time.time())[4]

second=time.localtime(time.time())[5]

hourAngle=-((hour+minute/60)/12)*2*np.pi+np.pi/2

minuteAngle=-((minute+second/60)/60)*2*np.pi+np.pi/2

secondAngle=-(second/60)*2*np.pi+np.pi/2

print(second)

hourHand.axis=vector(hourHandL*np.cos(hourAngle),hourHandL*np.sin(hourAngle),0)

minuteHand.axis=vector(minuteHandL*np.cos(minuteAngle),minuteHandL*np.sin(minuteAngle),0)

secondHand.axis=vector(secondHandL*np.cos(secondAngle),secondHandL*np.sin(secondAngle),0)

Artificial Intelligence, OpenCV

AI for Everyone: Parsing Hand Position Data in Mediapipe

November 4, 2021 admin

In this video lesson we show how to properly parse the hand position data coming from the Mediapipe Library. Mediapipe returns a rather complex data structure, and in this video we show you how to crack into that data structure, and pull out useful data; specifically, the (x,y) position data of the various landmark data on the hand. We create a python class that can be reused in future programs, allowing you to simply call the python class in order to properly pull out hand landmark data. Enjoy!

import cv2
print(cv2.__version__)

class mpHands:
    import mediapipe as mp
    def __init__(self,maxHands=2,tol1=.5,tol2=.5):
        self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)
    def Marks(self,frame):
        myHands=[]
        frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)
        results=self.hands.process(frameRGB)
        if results.multi_hand_landmarks != None:
            for handLandMarks in results.multi_hand_landmarks:
                myHand=[]
                for landMark in handLandMarks.landmark:
                    myHand.append((int(landMark.x*width),int(landMark.y*height)))
                myHands.append(myHand)
        return myHands

width=1280
height=720
cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)
cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)
cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)
cam.set(cv2.CAP_PROP_FPS, 30)
cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))
findHands=mpHands(2)
while True:
    ignore,  frame = cam.read()
    frame=cv2.resize(frame,(width,height))
    handData=findHands.Marks(frame)
    for hand in handData:
        for ind in [0,5,6,7,8]:
            cv2.circle(frame,hand[ind],25,(255,0,255),3)
    cv2.imshow('my WEBcam', frame)
    cv2.moveWindow('my WEBcam',0,0)
    if cv2.waitKey(1) & 0xff ==ord('q'):
        break
cam.release()

import cv2

print(cv2.__version__)

class mpHands:

import mediapipe as mp

def __init__(self,maxHands=2,tol1=.5,tol2=.5):

self.hands=self.mp.solutions.hands.Hands(False,maxHands,tol1,tol2)

def Marks(self,frame):

myHands=[]

frameRGB=cv2.cvtColor(frame,cv2.COLOR_BGR2RGB)

results=self.hands.process(frameRGB)

if results.multi_hand_landmarks != None:

for handLandMarks in results.multi_hand_landmarks:

myHand=[]

for landMark in handLandMarks.landmark:

myHand.append((int(landMark.x*width),int(landMark.y*height)))

myHands.append(myHand)

return myHands

width=1280

height=720

cam=cv2.VideoCapture(4,cv2.CAP_DSHOW)

cam.set(cv2.CAP_PROP_FRAME_WIDTH, width)

cam.set(cv2.CAP_PROP_FRAME_HEIGHT,height)

cam.set(cv2.CAP_PROP_FPS, 30)

cam.set(cv2.CAP_PROP_FOURCC,cv2.VideoWriter_fourcc(*'MJPG'))

findHands=mpHands(2)

while True:

ignore, frame = cam.read()

frame=cv2.resize(frame,(width,height))

handData=findHands.Marks(frame)

for hand in handData:

for ind in [0,5,6,7,8]:

cv2.circle(frame,hand[ind],25,(255,0,255),3)

cv2.imshow('my WEBcam', frame)

cv2.moveWindow('my WEBcam',0,0)

if cv2.waitKey(1) & 0xff ==ord('q'):

break

cam.release()

Technology Tutorials

AI For Everyone: Parsing Pose and Hand Data Landmarks from Mediapipe

Distinguish Between Right and Left Hands in MediaPipe

AI for Everyone: Developing a Gesture Based Arcade Game Based on Mediapipe and Python

Python 3D Graphics Tutorial 18: 3D Clock Model with Proper Hand Motion

AI for Everyone: Parsing Hand Position Data in Mediapipe

Making The World a Better Place One High Tech Project at a Time. Enjoy!