Category Archives: Python

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

November 25, 2021 admin

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()

OpenCV, Python

AI FOR EVERYONE HOMEWORK SOLUTION: Face Recognition with Live Video in OpenCV

October 23, 2021 admin

In this video lesson we show you how to recognize faces in live video from your WEBcam. It is the homework solution for lesson #16. This builds on the earlier video lessons in this series, so if any of this does not make sense, go back and check out some of the earlier lessons.

import cv2
import face_recognition as FR
font=cv2.FONT_HERSHEY_SIMPLEX
width=640
height=360
cam=cv2.VideoCapture(0,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'))

paulFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Paul McWhorter.jpg')
faceLoc=FR.face_locations(paulFace)[0]
paulFaceEncode=FR.face_encodings(paulFace)[0]

gavFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Gavriella Joy.jpg')
faceLoc=FR.face_locations(gavFace)[0]
gavFaceEncode=FR.face_encodings(gavFace)[0]

knownEncodings=[paulFaceEncode,gavFaceEncode]
names=['Paul McWhorter','Gavriella Joy']

while True:
    ignore,  unknownFace = cam.read()

    unknownFaceRGB=cv2.cvtColor(unknownFace,cv2.COLOR_RGB2BGR)
    faceLocations=FR.face_locations(unknownFaceRGB)
    unknownEncodings=FR.face_encodings(unknownFaceRGB,faceLocations)

    for faceLocation,unknownEncoding in zip(faceLocations,unknownEncodings):
        top,right,bottom,left=faceLocation
        print(faceLocation)
        cv2.rectangle(unknownFace,(left,top),(right,bottom),(255,0,0),3)
        name='Unknown Person'
        matches=FR.compare_faces(knownEncodings,unknownEncoding)
        print(matches)
        if True in matches:
            matchIndex=matches.index(True)
            print(matchIndex)
            print(names[matchIndex])
            name=names[matchIndex]
        cv2.putText(unknownFace,name,(left,top),font,.75,(0,0,255),2)

    cv2.imshow('My Faces',unknownFace)
    if cv2.waitKey(1) & 0xff ==ord('q'):
        break
cam.release()
cv2.destroyAllWindows()

import cv2

import face_recognition as FR

font=cv2.FONT_HERSHEY_SIMPLEX

width=640

height=360

cam=cv2.VideoCapture(0,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'))

paulFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Paul McWhorter.jpg')

faceLoc=FR.face_locations(paulFace)[0]

paulFaceEncode=FR.face_encodings(paulFace)[0]

gavFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Gavriella Joy.jpg')

faceLoc=FR.face_locations(gavFace)[0]

gavFaceEncode=FR.face_encodings(gavFace)[0]

knownEncodings=[paulFaceEncode,gavFaceEncode]

names=['Paul McWhorter','Gavriella Joy']

while True:

ignore, unknownFace = cam.read()

unknownFaceRGB=cv2.cvtColor(unknownFace,cv2.COLOR_RGB2BGR)

faceLocations=FR.face_locations(unknownFaceRGB)

unknownEncodings=FR.face_encodings(unknownFaceRGB,faceLocations)

for faceLocation,unknownEncoding in zip(faceLocations,unknownEncodings):

top,right,bottom,left=faceLocation

print(faceLocation)

cv2.rectangle(unknownFace,(left,top),(right,bottom),(255,0,0),3)

name='Unknown Person'

matches=FR.compare_faces(knownEncodings,unknownEncoding)

print(matches)

if True in matches:

matchIndex=matches.index(True)

print(matchIndex)

print(names[matchIndex])

name=names[matchIndex]

cv2.putText(unknownFace,name,(left,top),font,.75,(0,0,255),2)

cv2.imshow('My Faces',unknownFace)

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

break

cam.release()

cv2.destroyAllWindows()

OpenCV, Python

AI for Everyone LESSON 16: Face Recognition with OpenCV

October 19, 2021 admin

Guys in this lesson we learn how to not only detect faces in an image, but also recognize who the person is. We start by training our model with known faces, and then find and identify those faces in other pictures. For your convenience, I include the code developed in this lesson below.

import cv2
import face_recognition as FR
font=cv2.FONT_HERSHEY_SIMPLEX

donFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Donald Trump.jpg')
faceLoc=FR.face_locations(donFace)[0]
donFaceEncode=FR.face_encodings(donFace)[0]


nancyFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Nancy Pelosi.jpg')
faceLoc=FR.face_locations(nancyFace)[0]
nancyFaceEncode=FR.face_encodings(nancyFace)[0]

penceFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Mike Pence.jpg')
faceLoc=FR.face_locations(penceFace)[0]
penceFaceEncode=FR.face_encodings(penceFace)[0]

knownEncodings=[donFaceEncode,nancyFaceEncode,penceFaceEncode]
names=['Donald Trump','Nancy Pelosi','Mike Pence']

unknownFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/unknown/u1.jpg')
unknownFaceBGR=cv2.cvtColor(unknownFace,cv2.COLOR_RGB2BGR)
faceLocations=FR.face_locations(unknownFace)
unknownEncodings=FR.face_encodings(unknownFace,faceLocations)

for faceLocation,unknownEncoding in zip(faceLocations,unknownEncodings):
    top,right,bottom,left=faceLocation
    print(faceLocation)
    cv2.rectangle(unknownFaceBGR,(left,top),(right,bottom),(255,0,0),3)
    name='Unknown Person'
    matches=FR.compare_faces(knownEncodings,unknownEncoding)
    print(matches)
    if True in matches:
        matchIndex=matches.index(True)
        print(matchIndex)
        print(names[matchIndex])
        name=names[matchIndex]
    cv2.putText(unknownFaceBGR,name,(left,top),font,.75,(0,0,255),2)

cv2.imshow('My Faces',unknownFaceBGR)

cv2.waitKey(10000)

import cv2

import face_recognition as FR

font=cv2.FONT_HERSHEY_SIMPLEX

donFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Donald Trump.jpg')

faceLoc=FR.face_locations(donFace)[0]

donFaceEncode=FR.face_encodings(donFace)[0]

nancyFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Nancy Pelosi.jpg')

faceLoc=FR.face_locations(nancyFace)[0]

nancyFaceEncode=FR.face_encodings(nancyFace)[0]

penceFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/known/Mike Pence.jpg')

faceLoc=FR.face_locations(penceFace)[0]

penceFaceEncode=FR.face_encodings(penceFace)[0]

knownEncodings=[donFaceEncode,nancyFaceEncode,penceFaceEncode]

names=['Donald Trump','Nancy Pelosi','Mike Pence']

unknownFace=FR.load_image_file('C:/Users/Valued Customer/Documents/Python/demoImages/unknown/u1.jpg')

unknownFaceBGR=cv2.cvtColor(unknownFace,cv2.COLOR_RGB2BGR)

faceLocations=FR.face_locations(unknownFace)

unknownEncodings=FR.face_encodings(unknownFace,faceLocations)

for faceLocation,unknownEncoding in zip(faceLocations,unknownEncodings):

top,right,bottom,left=faceLocation

print(faceLocation)

cv2.rectangle(unknownFaceBGR,(left,top),(right,bottom),(255,0,0),3)

name='Unknown Person'

matches=FR.compare_faces(knownEncodings,unknownEncoding)

print(matches)

if True in matches:

matchIndex=matches.index(True)

print(matchIndex)

print(names[matchIndex])

name=names[matchIndex]

cv2.putText(unknownFaceBGR,name,(left,top),font,.75,(0,0,255),2)

cv2.imshow('My Faces',unknownFaceBGR)

cv2.waitKey(10000)

Technology Tutorials

Category Archives: Python

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

AI FOR EVERYONE HOMEWORK SOLUTION: Face Recognition with Live Video in OpenCV

AI for Everyone LESSON 16: Face Recognition with OpenCV

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