Hand Detection

Track hands in a live video feed with CVZone and count how many fingers are raised

Hand detection

Hand detection finds hands in images or video. MediaPipe goes further and finds 21 landmarks on each hand: the wrist and the joints and tips of every finger. That’s enough to recognise gestures, which makes hands a great way to control robots and games without touching anything.

The code

import cv2
from cvzone.HandTrackingModule import HandDetector

cap = cv2.VideoCapture(0)
detector = HandDetector(detectionCon=0.8, maxHands=2)

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

    hands, img = detector.findHands(img)

    if hands:
        hand = hands[0]
        fingers = detector.fingersUp(hand)
        print(hand["type"], "hand, fingers up:", fingers, "count:", fingers.count(1))

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

cap.release()
cv2.destroyAllWindows()

Watch the order

findHands() returns hands, img, the other way round from findFaces(). Mixing them up gives a confusing error the first time you use the result.

How it works

• HandDetector(detectionCon=0.8, maxHands=2) only accepts hands it’s at least 80% sure about, and tracks up to two
• detector.findHands(img) draws the hand skeleton and returns a list of hands plus the image
• Each hand is a dictionary with:
  • lmList - the 21 landmarks, each [x, y, z]. Landmark 0 is the wrist and 8 is the tip of the index finger
  • bbox and center - the box around the hand
  • type - "Left" or "Right"
• detector.fingersUp(hand) returns five numbers, thumb first, with 1 for a raised finger and 0 for a folded one. So [0, 1, 1, 0, 0] is a peace sign

Measuring a pinch

The distance between two landmarks makes a handy control. Add this inside the if hands: block:

        thumb_tip = hand["lmList"][4][0:2]
        index_tip = hand["lmList"][8][0:2]
        length, info, img = detector.findDistance(thumb_tip, index_tip, img)
        print("Pinch distance:", int(length))

findDistance() draws a line between the two points and returns the distance in pixels. Pinch your fingers together and apart to watch it change. You could use this as a volume knob or to set a robot’s speed.

Try it yourself

Show a number from 0 to 5 with your fingers and print it in big letters. Then make the program print “Rock”, “Paper” or “Scissors” from your hand shape.

< Previous Next >

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