Annotating Your Video

Overlay live text and feedback on your video with CVZone's putTextRect, and build a curl counter

Annotating your video

Printing numbers to the terminal is fine for testing, but it’s much nicer to show feedback right on the video. CVZone’s putTextRect() draws text on a coloured background, so it’s easy to read whatever is behind it.

Adding text

import cvzone

img, bbox = cvzone.putTextRect(img, "Hello!", (50, 50))

• The position is where the text starts, in pixels from the top left corner
• It returns the image and the box around the text

You can change the look with a few options:

img, _ = cvzone.putTextRect(img, "Hello!", (50, 50),
                            scale=2, thickness=2,
                            colorT=(255, 255, 255),   # text colour
                            colorR=(255, 0, 255))     # background colour

Colours are BGR

OpenCV colours are (blue, green, red), not (red, green, blue). So (0, 0, 255) is red and (255, 0, 0) is blue.

A bicep curl counter

Let’s put the last lesson to work. This program watches your left elbow and counts a curl each time you bend your arm fully and straighten it again.

import cv2
import cvzone
from cvzone.PoseModule import PoseDetector

cap = cv2.VideoCapture(0)
detector = PoseDetector()

count = 0
arm_bent = False

while True:
    success, img = cap.read()
    if not success:
        break

    img = detector.findPose(img)
    lmList, _ = detector.findPosition(img, draw=False)

    if lmList:
        angle, img = detector.findAngle(lmList[11][0:2], lmList[13][0:2], lmList[15][0:2], img=img)

        if angle < 50 and not arm_bent:
            arm_bent = True
        elif angle > 150 and arm_bent:
            arm_bent = False
            count += 1

        status = "Bent" if arm_bent else "Straight"
        colour = (0, 165, 255) if arm_bent else (0, 200, 0)
        img, _ = cvzone.putTextRect(img, f"Angle: {int(angle)}", (50, 60), scale=2, thickness=2)
        img, _ = cvzone.putTextRect(img, status, (50, 120), scale=2, thickness=2, colorR=colour)

    img, _ = cvzone.putTextRect(img, f"Curls: {count}", (50, 200), scale=3, thickness=3, colorR=(255, 0, 0))

    cv2.imshow("Curl Counter", img)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

How it works

• arm_bent remembers whether the arm was last seen bent or straight
• A curl only counts when the arm goes from bent (below 50 degrees) back to straight (above 150 degrees). The gap between the two numbers stops one wobbly movement being counted several times
• The status label changes colour so you can see the state at a glance
• The count is drawn outside the if lmList: block, so it stays on screen even when you step out of view

No degree symbol

OpenCV’s built-in fonts can’t draw °, so it would show up as ??. Stick to plain text in your labels.

Further exploration

• Both arms: count curls on the right arm too, using landmarks 12, 14 and 16
• Other exercises: squats use the hip, knee and ankle (23, 25, 27)
• Other drawing tools: cvzone.cornerRect() draws a stylish box around a bbox, and cvzone.stackImages() puts several images side by side

What’s next?

• Building Object Detection Models with Raspberry Pi AI Camera - train and run your own object detection model
• Combine hand or face tracking with a robot from Build a SMARS Robot or BurgerBot

< Previous

You can use the arrows  ← → on your keyboard to navigate between lessons.