suhaasborra.com

genx · agent runtime · hybrid rag
directory

bloghdai-ai

hdai

Your webcam is already a sensor. HD-AI makes it a touchless controller.

PalmRecognitionPublicblog post

the idea

the problem

Touchscreens assume clean hands and an arm's length of space. Kiosks, labs, kitchens and factory floors break both assumptions. Depth cameras fix that, but they are one more device to buy, mount and calibrate.

the approach

HD-AI reads hand pose from an ordinary camera and maps gestures to input events. Palm detection finds the hand, landmarks follow it frame by frame, and a thin gesture layer emits the same events a mouse would. No extra hardware, no touch.

what's done

  • Palm detection plus landmark tracking, running from a single webcam feed.

  • Gesture-to-event mapping: point, pinch, move, click.

  • A reference app that turns the stream into a pointer you can actually use.

what's missing

  • Packaging. Right now you clone a repository; you do not install it.

  • A calibration screen for different cameras and room lighting.

  • Latency work on low-end machines. The model is happier with a GPU than without one.

how to try it

git clone https://github.com/Suhaas-code/PalmRecognitionPublic
cd PalmRecognitionPublic
pip install -r requirements.txt
python app.py

A webcam and Python 3.10+ is enough to start. Source: PalmRecognitionPublic.

all posts