Real-Time Finger Counting Using Python, OpenCV & MediaPipe

Ever wanted to control something on your screen just by holding up fingers? This project turns any webcam into a simple gesture-recognition tool using computer vision — a great entry point into AI/ML for anyone coming from an embedded systems or Arduino background.

The project uses MediaPipe Hands, a pre-trained hand-tracking model from Google, to detect 21 landmark points on a hand in every video frame — knuckles, fingertips, and the wrist. OpenCV handles capturing the webcam feed and drawing the results on screen.

The finger-counting logic itself is simple geometry: for four of the five fingers, if the fingertip landmark sits above the middle-joint landmark in the frame, that finger is counted as "up." The thumb is a special case since it moves sideways rather than up-down, so it's checked by comparing X-coordinates instead of Y.

The output window shows:

• A live skeleton overlay of your detected hand
• A large on-screen number counting fingers from 0 to 5
• Real-time FPS count
• Detected hand label (Left / Right)

This is a solid beginner AI/ML project because it introduces:

• Real-time video processing with OpenCV
• Using a pre-trained ML model (MediaPipe) without training anything yourself
• Basic landmark/geometry-based logic for gesture recognition

Where to take it next: this finger count can become an input trigger for other projects — for example, sending a gesture command over serial or Wi-Fi to an ESP32 to control an LED, servo, or relay. That's a natural bridge between this AI/ML tutorial and your embedded systems projects.