Skip to content

Victor Murcia

• About
• Engineering
• Research
• Publications
• Writing
• Archive

Aug 17, 2022 · Archived

Real Time Facial Recognition in Python

Adding filters to live video data

From earlier work — part of my archive of foundational machine-learning & data-science projects.

Real Time Face Tracking in Python

This is a quick post showcasing how to write a real-time face/eye tracking routine in Python using OpenCV. I'll show how this routine can be used to add fun filters :]

Modules Used in this work

Here are the most important modules used for this project:

• open cv : Used to carry out operations on images

The API documentation for each of these modules can be found here:

• https://docs.opencv.org/3.4/index.html

#Import modules
import cv2

def objectTracker1(cascPath):
    
    faceCascade = cv2.CascadeClassifier(cascPath)
    
    cv2.namedWindow("preview")
    vc = cv2.VideoCapture(0)
    
    while True: # try to get the first frame
        rval, frame = vc.read()
        # Capture frame-by-frame
        ret, frame = vc.read()
        
        #Convert frame to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    
        objects = faceCascade.detectMultiScale(gray, scaleFactor=1.2, 
                                               minNeighbors=1, minSize=(40, 40), 
                                               flags=cv2.CASCADE_SCALE_IMAGE)
        # Draw a rectangle around the faces
        for (x, y, w, h) in objects:
            cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)
        
        cv2.imshow("preview", frame)
        rval, frame = vc.read()
        key = cv2.waitKey(20)
        if key == 27: # exit on ESC
            break

    vc.release()
    cv2.destroyWindow("preview")

In my previous post where I determined player jersey colors from video footage, I briefly explored using the HOG package to detect human figures where I had initial success. Upon looking through the various classifiers in OpenCVs documentation, I came across the 'Haar Cascades' libraries which extend the base HOG descriptors to include things like cats and different facial profiles.

After playing with the cascade parameters a bit, I got the result below

face_cascade_Path = r'haarcascades\haarcascade_frontalface_alt.xml'
objectTracker1(face_cascade_Path)

So long as I don't move too quickly the detector does a pretty good job at solely detecting my face! Let me try other cascades and see how that works :]

eye_cascade_Path  = r'haarcascades\haarcascade_eye.xml'
objectTracker1(eye_cascade_Path)

Cool! I can detect faces and eyes in real time using these cascades. However, you can see that there are a several smaller squares appearing outside my face that are clearly false positives. Can I combine them together though? Let me try to restrict the "eye detections" to occur only within the facial bounding box...

def objectTracker2(facePath, eyePath):
    
    faceCascade = cv2.CascadeClassifier(facePath)
    eyeCascade  = cv2.CascadeClassifier(eyePath)
    cv2.namedWindow("preview")
    vc = cv2.VideoCapture(0)
    
    while True: # try to get the first frame
        rval, frame = vc.read()
        # Capture frame-by-frame
        ret, frame = vc.read()
        
        #Convert frame to grayscale
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    
        faces = faceCascade.detectMultiScale(gray, scaleFactor=1.2, 
                                               minNeighbors=1, minSize=(40, 40), 
                                               flags=cv2.CASCADE_SCALE_IMAGE)
        # Draw a rectangle around the faces
        for (x, y, w, h) in faces:
            cv2.rectangle(frame, (x, y), (x+w, y+h), (0, 255, 0), 2)
            #select face as region of interest 
            roi_g =  gray[y:y+h,x:x+h]
            roi_c = frame[y:y+h,x:x+h]
            #within region of interest find eyes
            eyes = eyeCascade.detectMultiScale(roi_g)
            #for each eye
            for (ex,ey,ew,eh) in eyes:
                #draw retangle around eye
                cv2.rectangle(roi_c, (ex,ey),(ex+ew,ey+eh),(0,255,0),2)
                    
        cv2.imshow("preview", frame)
        rval, frame = vc.read()
        key = cv2.waitKey(20)
        if key == 27: # exit on ESC
            break

    vc.release()
    cv2.destroyWindow("preview")

face_cascade_Path = r'haarcascades\haarcascade_frontalface_alt.xml'
eye_cascade_Path  = r'haarcascades\haarcascade_eye.xml'

objectTracker2(face_cascade_Path, eye_cascade_Path)

This worked pretty good! Now we can start adding all sorts of funky stuff onto our process! What's important here is where we want to position our added thing. For instance, if wanted to place a hat on top of our head then we'd need to determine the top coordinates of our head, if we wanted to add glasses on top of our eyes then we'd need those coordinates, and so on and so forth.

---

← Back to the archive

Victor Murcia, PhD AI Engineering · Scientific Computing · Technical Leadership

GitHub LinkedIn arXiv RSS

© 2026 Victor Murcia