tarikurrahmanbd/control_pc_using_hand-gesture-main

This project lets you control your computer's mouse using hand gestures picked up by a webcam, no extra hardware needed beyond the camera. It is built in Python using OpenCV for video handling and MediaPipe for detecting hand landmarks, and this is version 3, which reorganizes the code into a clean set of modules. The system understands a handful of specific hand shapes. Raising only your index finger moves the cursor to follow your fingertip. Pinching your thumb and index finger together, and holding the pinch for a quarter of a second, triggers a left click, while pinching thumb and middle finger triggers a right click. Raising your index and middle fingers together, like a peace sign, switches into scroll mode, where moving your hand up or down scrolls the page in that direction. Making a closed fist pauses all control instantly, turning the screen border red as a visual warning, and pressing Q on the keyboard shuts the whole program down as a safety measure. Beyond gesture recognition, the project adds cursor inertia and acceleration so the pointer feels closer to a normal mouse instead of jumping around, along with filtering to smooth out the natural shake of a hand held in front of a webcam. It also automatically pauses itself if it loses track of your hand or if the frame rate drops too low, and it uses short cooldowns and pattern checks to avoid triggering clicks by accident. An on-screen overlay shows the current mode, frame rate, and detected gesture while you use it. To run it, you need Python 3.8 or newer and a webcam. You install the dependencies from requirements.txt, download a MediaPipe hand landmarker model file (a link is provided, or you can fetch it with a PowerShell command), and then run main.py. A few settings like scroll speed, cooldown timing, and cursor smoothing can be adjusted in config.py. The author describes it as a project built for portfolio purposes, with voice control and an air-typing keyboard listed as possible future additions.