The repository contains 8 mini-projects which are built upon majorly two libraries MediaPipe and OpenCV. While MediaPipe offers a number of functionalities to perform hand tracking, pose estimation etc, OpenCV helps users to interact with these models.
The amazing thing about all of these projects is that they run on CPU in real-time!
Detecting and tracking 21 3D hand landmarks at around 25 frames per second. To see the code go here.
Finding and tracking 33 3D full-body landmarks from an image or video. To see the code go here.
Recognize 6 face landmarks. To see the code go here.
Spots 468 3D face landmarks in real-time. To see the code go [here]face_mesh).
Uses the module from hand tracking and applies a little math to count the number of fingers. This may come in handy when you want your computer to do some tasks depending upon the number of fingers you are holding up. To see the code go here.
Here we employ a pose estimation module to track movements of the body. To see the code go here.
We create a canvas to select a tool to paint, then with some maths and hand tracking module, we fill color in our canvas. To see the code go here.
I was the most excited for this project, where after waving your hand in the air you can play music or pretty much do any task. To see the code go here.