Tag Archives: openCV

AI on the Edge LESSON 18: Display Frames Per Second (FPS) on openCV Video Window

May 27, 2026 admin

In today’s lesson, we add a clean, real-time Frames Per Second (FPS) counter directly onto our live OpenCV video window. Displaying FPS on screen is an essential tool for anyone working with camera-based AI projects on the Raspberry Pi. It gives you immediate feedback on your actual processing performance, helps with optimization, and makes your projects look more professional and polished.

In this lesson, we configure the Picamera2 library to run at 1280×720 resolution with a target of 60 frames per second. We then implement a smoothed FPS calculation using a weighted rolling average, which prevents the displayed value from jumping around wildly. Finally, we overlay the FPS text in the lower-left corner of the video frame using OpenCV’s putText() function, with font size and thickness that scale appropriately with the resolution.

This technique forms an important foundation for future lessons, as we will continue adding more information and graphics directly onto the live video stream. Understanding how to efficiently display performance metrics is key to developing responsive and practical edge AI applications.

In this lesson, this is the code which we develop:

import cv2
import time
from picamera2 import Picamera2
piCam = Picamera2()
W=1280
H=720
tStart = time.time()
fps = 0
RES = (W,H)
piCam.preview_configuration.main.size = RES
piCam.preview_configuration.main.format = "RGB888"
piCam.preview_configuration.controls.FrameRate=60
piCam.preview_configuration.align()
piCam.configure("preview")
piCam.start()

textLowerLeft = (int(W*.01),int(H*.05))
fontFace = cv2.FONT_HERSHEY_SIMPLEX
fontThickness = int(W/425)
fontScale = H*.0015
fontColor = (0,0,255)

while True:
    deltaT = time.time() - tStart
    tStart=time.time()
    fps = fps*.95 + (1/deltaT)*.05
    frame= piCam.capture_array()
    frame=cv2.flip(frame,-1)
    myText = "FPS: "+str(round(fps,1))
    cv2.putText(frame,myText,textLowerLeft,fontFace,fontScale,fontColor,fontThickness)
    cv2.imshow("Camera", frame)
    cv2.moveWindow("Camera",0,60)
    if cv2.waitKey(1)==ord('q'):
        break
cv2.destroyAllWindows()
print('Program Terminated')

import cv2

import time

from picamera2 import Picamera2

piCam = Picamera2()

W=1280

H=720

tStart = time.time()

fps = 0

RES = (W,H)

piCam.preview_configuration.main.size = RES

piCam.preview_configuration.main.format = "RGB888"

piCam.preview_configuration.controls.FrameRate=60

piCam.preview_configuration.align()

piCam.configure("preview")

piCam.start()

textLowerLeft = (int(W*.01),int(H*.05))

fontFace = cv2.FONT_HERSHEY_SIMPLEX

fontThickness = int(W/425)

fontScale = H*.0015

fontColor = (0,0,255)

while True:

deltaT = time.time() - tStart

tStart=time.time()

fps = fps*.95 + (1/deltaT)*.05

frame= piCam.capture_array()

frame=cv2.flip(frame,-1)

myText = "FPS: "+str(round(fps,1))

cv2.putText(frame,myText,textLowerLeft,fontFace,fontScale,fontColor,fontThickness)

cv2.imshow("Camera", frame)

cv2.moveWindow("Camera",0,60)

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

break

cv2.destroyAllWindows()

print('Program Terminated')

AI On the Edge

AI on the Edge LESSON 17: Decorating and Annotating Video Frames in openCV

May 22, 2026 admin

Welcome to AI on the Edge – Lesson 17: Decorating and Annotating Video Frames in OpenCV. In this lesson we take our live video stream from the Raspberry Pi camera and start making it really useful and professional-looking. Now that we can grab frames and display them, it’s time to learn how to draw directly on top of those frames. We’re talking rectangles, lines, arrows, circles, and crisp text overlays — all the visual elements you’ll need when you start adding real AI like face detection or object recognition.

You’ll see exactly how to use OpenCV’s drawing functions to create clean, scalable annotations that look great whether you’re running at 320×180 for maximum speed or higher resolutions like 1280×720. We cover how to control line thickness, use filled shapes, position text properly, and most importantly, how to make all your drawings scale automatically with your chosen resolution so everything stays nicely proportioned.

By the end of this lesson you’ll have the skills to draw bounding boxes around detected objects, add confidence scores, label people or items, draw tracking lines — basically anything you need to show what your AI is seeing. This is one of those foundational skills that you’ll use over and over again in your computer vision projects.As always, I encourage you to type the code along with me in the video, then start playing with colors, sizes, positions, and messages. Change things around, break it, and make it your own. That’s the best way to really learn this stuff.

So fire up your Raspberry Pi 5, grab that camera, and let’s start turning raw video frames into clear, informative, and great-looking annotated output!

import cv2
from picamera2 import Picamera2
piCam = Picamera2()
W=1280
H=720
# W= 640
# H= 360
W=320
H=180
RES = (W,H)
piCam.preview_configuration.main.size = RES
piCam.preview_configuration.main.format = "RGB888"
piCam.preview_configuration.controls.FrameRate=60
piCam.preview_configuration.align()
piCam.configure("preview")
piCam.start()
while True:
    frame= piCam.capture_array()
    frame=cv2.flip(frame,-1)

    upperLeft = (int((W-1)*.1),int((H-1)*.1))
    lowerRight = (int((W-1)*.9),int((H-1)*.9))
    myColor = (255,0,0)
    myThick = int(W/250)
    cv2.rectangle(frame,upperLeft,lowerRight,myColor,myThick)
    
    upperLeft = (int((W-1)*.3),int((H-1)*.3))
    lowerRight = (int((W-1)*.7),int((H-1)*.7))
    myColor = (0,0,255)
    myThick = int(W/250)
    myThick = -1
    cv2.rectangle(frame,upperLeft,lowerRight,myColor,myThick)
 
    startingPoint = (int((W-1)*.1),int((H-1)*.1))
    endingPoint = (int((W-1)*.9),int((H-1)*.9))
    myColor = (255,0,0)
    myThick = int(W/250)
    cv2.line(frame,startingPoint,endingPoint,myColor,myThick)
 
    startingPoint = (int((W-1)*.7),int((H-1)*.3))
    endingPoint = (int((W-1)*.3),int((H-1)*.7))
    myColor = (0,255,0)
    myThick = int(W/250)
    cv2.arrowedLine(frame,startingPoint,endingPoint,myColor,myThick)
 
    myCenter = (int((W-1)/2),int((H-1)/2))
    myRadius = int(.1*(H-1))
    myThick = int(W/250)
    myColor = (0,255,0)
    cv2.circle(frame,myCenter,myRadius,myColor,myThick)
    
    myCenter1 = (int((W-1)*.3),int((H-1)*.3))
    myCenter2 = (int((W-1)*.7),int((H-1)*.7))
    myCenter3 = (int((W-1)*.7),int((H-1)*.3))
    myCenter4 = (int((W-1)*.3),int((H-1)*.7))
    myRadius = int((W-1)*.05)
    myColor = (255,0,0)
    myThick= -1
    cv2.circle(frame,myCenter1,myRadius,myColor,myThick)
    cv2.circle(frame,myCenter2,myRadius,myColor,myThick)
    cv2.circle(frame,myCenter3,myRadius,myColor,myThick)
    cv2.circle(frame,myCenter4,myRadius,myColor,myThick)  

    myText = "My Excellent Label"
    textLowerLeft= (int(W*.01),int(H*.05))
    fontFace = cv2.FONT_HERSHEY_SIMPLEX
    fontThickness = int(W/450)
    fontScale = H*.0015
    fontColor = (255,0,255)
    cv2.putText(frame,myText,textLowerLeft,fontFace,fontScale, fontColor, fontThickness)


    cv2.imshow("Camera", frame)
    

    cv2.moveWindow("Camera",0,60)
    if cv2.waitKey(1)==ord('q'):
        break
cv2.destroyAllWindows()
print('Program Terminated')

import cv2

from picamera2 import Picamera2

piCam = Picamera2()

W=1280

H=720

# W= 640

# H= 360

W=320

H=180

RES = (W,H)

piCam.preview_configuration.main.size = RES

piCam.preview_configuration.main.format = "RGB888"

piCam.preview_configuration.controls.FrameRate=60

piCam.preview_configuration.align()

piCam.configure("preview")

piCam.start()

while True:

frame= piCam.capture_array()

frame=cv2.flip(frame,-1)

upperLeft = (int((W-1)*.1),int((H-1)*.1))

lowerRight = (int((W-1)*.9),int((H-1)*.9))

myColor = (255,0,0)

myThick = int(W/250)

cv2.rectangle(frame,upperLeft,lowerRight,myColor,myThick)

upperLeft = (int((W-1)*.3),int((H-1)*.3))

lowerRight = (int((W-1)*.7),int((H-1)*.7))

myColor = (0,0,255)

myThick = int(W/250)

myThick = -1

cv2.rectangle(frame,upperLeft,lowerRight,myColor,myThick)

startingPoint = (int((W-1)*.1),int((H-1)*.1))

endingPoint = (int((W-1)*.9),int((H-1)*.9))

myColor = (255,0,0)

myThick = int(W/250)

cv2.line(frame,startingPoint,endingPoint,myColor,myThick)

startingPoint = (int((W-1)*.7),int((H-1)*.3))

endingPoint = (int((W-1)*.3),int((H-1)*.7))

myColor = (0,255,0)

myThick = int(W/250)

cv2.arrowedLine(frame,startingPoint,endingPoint,myColor,myThick)

myCenter = (int((W-1)/2),int((H-1)/2))

myRadius = int(.1*(H-1))

myThick = int(W/250)

myColor = (0,255,0)

cv2.circle(frame,myCenter,myRadius,myColor,myThick)

myCenter1 = (int((W-1)*.3),int((H-1)*.3))

myCenter2 = (int((W-1)*.7),int((H-1)*.7))

myCenter3 = (int((W-1)*.7),int((H-1)*.3))

myCenter4 = (int((W-1)*.3),int((H-1)*.7))

myRadius = int((W-1)*.05)

myColor = (255,0,0)

myThick= -1

cv2.circle(frame,myCenter1,myRadius,myColor,myThick)

cv2.circle(frame,myCenter2,myRadius,myColor,myThick)

cv2.circle(frame,myCenter3,myRadius,myColor,myThick)

cv2.circle(frame,myCenter4,myRadius,myColor,myThick)

myText = "My Excellent Label"

textLowerLeft= (int(W*.01),int(H*.05))

fontFace = cv2.FONT_HERSHEY_SIMPLEX

fontThickness = int(W/450)

fontScale = H*.0015

fontColor = (255,0,255)

cv2.putText(frame,myText,textLowerLeft,fontFace,fontScale, fontColor, fontThickness)

cv2.imshow("Camera", frame)

cv2.moveWindow("Camera",0,60)

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

break

cv2.destroyAllWindows()

print('Program Terminated')

AI On the Edge, Tutorial

AI on the Edge LESSON 16: Control Pan/Tilt Camera Position Using Voice Commands

May 20, 2026 admin

In AI on the Edge Lesson 16, we take a big step forward by combining voice recognition with physical motion. In this project, you will build a voice-controlled pan/tilt camera system. Using simple spoken commands such as “right,” “left,” “up,” “down,” and “quit,” you can move the Raspberry Pi camera in real time. This lesson brings together the Fusion HAT+ servo control, the Speech-to-Text (STT) capabilities we explored earlier, live video streaming with picamera2 and OpenCV, and multithreading to keep everything running smoothly.

The hardware setup is straightforward. We connect two servos to the Fusion HAT+ — one for pan (horizontal movement) on pin 2 and one for tilt (vertical movement) on pin 3. The Raspberry Pi Camera is mounted on a pan/tilt mechanism so it can physically follow your voice commands. We start the camera at a neutral position (pan = 0°, tilt = -20°) and define step sizes so the movement feels responsive but controlled.

The Python code uses two main threads: one for continuous voice listening and another for displaying the live video feed. In the listening thread, we create an STT object and continuously wait for voice input. When a command is recognized, we adjust the pan or tilt angle accordingly and immediately send the new position to the appropriate servo. The main loop captures frames from the Pi Camera, flips them for correct orientation, displays them in an OpenCV window, and checks for the ‘q’ key to exit gracefully.

This project demonstrates several important concepts working together: real-time voice command processing, servo motor control, camera streaming with picamera2 at 1280×720 resolution and 60 fps, and proper use of threading so that listening and video display do not block each other. You will also notice how we use global variables carefully to share the current pan and tilt positions between the threads.

By the end of this lesson, you will have a working voice-controlled camera that you can point anywhere you want just by talking to it. This is an excellent foundation for more advanced projects such as voice-controlled object tracking, security cameras, or interactive AI assistants that can both see and move.The complete code is provided below, along with explanations of the key sections. Feel free to experiment with different step sizes (xDelta and yDelta), starting angles, or even add new voice commands once you are comfortable with the basic version.

This is the code developed in the video lesson:

from fusion_hat.servo import Servo
from fusion_hat.stt import STT
import threading
import cv2
from picamera2 import Picamera2
piCam = Picamera2()

panPin = 2
tiltPin = 3

panServo = Servo(panPin)
tiltServo = Servo(tiltPin)

x = 0
y =-20

xDelta = 10
yDelta = 5

panServo.angle(x)
tiltServo.angle(y)

W=1280
H=720
RES = (W,H)
piCam.preview_configuration.main.size = RES
piCam.preview_configuration.main.format = "RGB888"
piCam.preview_configuration.controls.FrameRate=60
piCam.preview_configuration.align()
piCam.configure("preview")
piCam.start()

running = True
    
def listenThread():
    global running, x, y
    stt = STT(language = "en-us")
    while running:
        print("Listening . . .")
        result = stt.listen(stream = False)
        command = result.strip()
        print("Command was: ", command)
        if command == "right":
            x = x + xDelta
            panServo.angle(x)
        if command == "left":
            x= x - xDelta
            panServo.angle(x)
        if command == "up":
            y = y - yDelta
            tiltServo.angle(y)
        if command == "down":
            y = y + yDelta
            tiltServo.angle(y)
        if command == "quit":
            running = False
            break
    print("Thread Terminated")
myThread = threading.Thread(target = listenThread, daemon = True)
myThread.start()
while running:
    frame= piCam.capture_array()
    frame=cv2.flip(frame,-1)
    cv2.imshow("Camera", frame)
    cv2.moveWindow("Camera",0,60)
    if cv2.waitKey(1)==ord('q'):
        running = False
        break
cv2.destroyAllWindows()
piCam.stop()
print('Program Terminated')

from fusion_hat.servo import Servo

from fusion_hat.stt import STT

import threading

import cv2

from picamera2 import Picamera2

piCam = Picamera2()

panPin = 2

tiltPin = 3

panServo = Servo(panPin)

tiltServo = Servo(tiltPin)

x = 0

y =-20

xDelta = 10

yDelta = 5

panServo.angle(x)

tiltServo.angle(y)

W=1280

H=720

RES = (W,H)

piCam.preview_configuration.main.size = RES

piCam.preview_configuration.main.format = "RGB888"

piCam.preview_configuration.controls.FrameRate=60

piCam.preview_configuration.align()

piCam.configure("preview")

piCam.start()

running = True

def listenThread():

global running, x, y

stt = STT(language = "en-us")

while running:

print("Listening . . .")

result = stt.listen(stream = False)

command = result.strip()

print("Command was: ", command)

if command == "right":

x = x + xDelta

panServo.angle(x)

if command == "left":

x= x - xDelta

panServo.angle(x)

if command == "up":

y = y - yDelta

tiltServo.angle(y)

if command == "down":

y = y + yDelta

tiltServo.angle(y)

if command == "quit":

running = False

break

print("Thread Terminated")

myThread = threading.Thread(target = listenThread, daemon = True)

myThread.start()

while running:

frame= piCam.capture_array()

frame=cv2.flip(frame,-1)

cv2.imshow("Camera", frame)

cv2.moveWindow("Camera",0,60)

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

running = False

break

cv2.destroyAllWindows()

piCam.stop()

print('Program Terminated')

AI On the Edge

AI on the Edge LESSON 15: Use the Raspberry Pi Camera in openCV to Create Live Video

May 15, 2026 admin

Hey everyone, welcome to Lesson 15 of the AI on the Edge series!

In today’s lesson, we take a very important step forward. We finally bring the Raspberry Pi Camera into our OpenCV world so we can capture live video and start building real computer vision projects. Today we learn how to pull live frames directly from the official Raspberry Pi camera using picamera2 and display them smoothly with OpenCV.

This lesson is all about building a clean, reliable foundation. I walk you through how to properly configure the Pi Camera with the modern picamera2 library — setting the resolution to 1280×720, choosing the right format, and pushing the frame rate up to 60fps. Then we bring those frames straight into OpenCV so we can see live video in a window. You’ll also learn why we use RGB888 format and how to organize your code so it stays clean as our projects get more complex.

Getting reliable live video from the Pi Camera is one of those foundational skills that opens the door to everything we’re going to do in this class — face detection, object tracking, color tracking, motion detection, and all the exciting AI projects still ahead. Once you have solid camera access, the real fun begins.

I kept this lesson straightforward on purpose. I want you to have a rock-solid base that you can build upon without fighting technical problems later. By the end of this video, you’ll have a clean, responsive live video stream running from your Raspberry Pi Camera, ready for all the computer vision magic we’re about to add in the coming lessons.

So fire up your Raspberry Pi, grab your camera module, and let’s get that live video rolling! As always, I encourage you to type the code along with me and experiment with it. Change the resolution, try different frame rates, and make it your own.

Are you ready? Let’s dive in!

In today’s lesson, this is the code which we developed:

import cv2
from picamera2 import Picamera2
piCam = Picamera2()
W=1280
H=720
RES = (W,H)
piCam.preview_configuration.main.size = RES
piCam.preview_configuration.main.format = "RGB888"
piCam.preview_configuration.controls.FrameRate=60
piCam.preview_configuration.align()
piCam.configure("preview")
piCam.start()
while True:
    frame= piCam.capture_array()
    #frame=cv2.flip(frame,-1)
    cv2.imshow("Camera", frame)
    cv2.moveWindow("Camera",0,60)
    if cv2.waitKey(1)==ord('q'):
        break
cv2.destroyAllWindows()
print('Program Terminated')

import cv2

from picamera2 import Picamera2

piCam = Picamera2()

W=1280

H=720

RES = (W,H)

piCam.preview_configuration.main.size = RES

piCam.preview_configuration.main.format = "RGB888"

piCam.preview_configuration.controls.FrameRate=60

piCam.preview_configuration.align()

piCam.configure("preview")

piCam.start()

while True:

frame= piCam.capture_array()

#frame=cv2.flip(frame,-1)

cv2.imshow("Camera", frame)

cv2.moveWindow("Camera",0,60)

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

break

cv2.destroyAllWindows()

print('Program Terminated')

This is the schematic we are using in these lessons:

Fusion Hat Circuit Diagram — This is the circuit we will use moving forward in the class

AI On the Edge, Tutorial

AI on the Edge LESSON 7: Homework Solution for Dimmable LED

April 18, 2026 admin

In Lesson 6, I gave you a homework challenge: build a dimmable LED using a potentiometer. In today’s Lesson 7, we go through the solution together step-by-step.

This lesson is all about taking analog input from a potentiometer and converting it into smooth PWM output to control the brightness of an LED. It’s a very practical project because it teaches you how to read real-world analog values and turn them into useful control signals — skills we’ll use again and again as we build smarter AI-powered projects.

In the video, I walk you through the complete working code. You’ll see how we read the potentiometer value (0 to 4095), convert that raw number into a proper brightness percentage using a bit of math (with a nice logarithmic curve so the brightness feels natural to the human eye), and then send that value to the LED using PWM. The result is a very smooth, responsive dimmer that feels professional.

Even though this seems like a simple project, it’s actually an important stepping stone. Understanding how to read sensors and smoothly control outputs is fundamental to building real AI on the Edge systems — whether you’re controlling motors, adjusting screen brightness, or varying the speed of a robot based on sensor input.

By the end of this lesson, you should have a solid understanding of how to combine the ADC (Analog to Digital Converter) with PWM output, and more importantly, how to think about mapping real-world inputs to useful outputs.

So if you did the homework, great job! If you got stuck, don’t worry — we go through the full solution together. And as always, I strongly encourage you to take the code and make it your own. Try changing the response curve, add multiple LEDs with different colors, or combine it with things we’ve learned in earlier lessons.

This is the kind of foundational hardware skill that will serve you well as we continue moving deeper into the AI on the Edge class. You’re doing great — keep going!

We are still using the schematic from our earlier project.

In this lesson, this is the code which we came up with:

#!/usr/bin/env python3
from fusion_hat.adc import ADC
from fusion_hat.pwm import PWM
from time import sleep
import math
potPin=0
myPot = ADC(0)
redPin=5
redLED=PWM(redPin)

while True:
    potVal = myPot.read()                     # 0 to 4095
    writeVal=100**(potVal/4095)-.005
    redLED.pulse_width_percent(int(writeVal))
    print(potVal,writeVal)
    sleep(0.1)

#!/usr/bin/env python3

from fusion_hat.adc import ADC

from fusion_hat.pwm import PWM

from time import sleep

import math

potPin=0

myPot = ADC(0)

redPin=5

redLED=PWM(redPin)

while True:

potVal = myPot.read() # 0 to 4095

writeVal=100**(potVal/4095)-.005

redLED.pulse_width_percent(int(writeVal))

print(potVal,writeVal)

sleep(0.1)

Technology Tutorials

Tag Archives: openCV

AI on the Edge LESSON 18: Display Frames Per Second (FPS) on openCV Video Window

AI on the Edge LESSON 17: Decorating and Annotating Video Frames in openCV

AI on the Edge LESSON 16: Control Pan/Tilt Camera Position Using Voice Commands

AI on the Edge LESSON 15: Use the Raspberry Pi Camera in openCV to Create Live Video

AI on the Edge LESSON 7: Homework Solution for Dimmable LED

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