Awesome

HandPoseBarracuda

HandPoseBarracuda is a proof-of-concept implementation of a neural network hand/finger tracker that works with a monocular color camera.

Basically, HandPoseBarracuda is a partial port of the MediaPipe Hands pipeline. Although it is not a straight port of the original package, it uses the same basic design and the same pre-trained models.

Note that this is just a proof-of-concept implementation. It lacks some essential features needed for practical applications:

• It only accepts a single hand. Although you can reuse the most part of the implementation, you will need to redesign the system to support multiple hands.
• It only supports screen-space (2D) positions and relative depths from a palm. You will need to implement a screen-to-world-space projector for 3D applications.

Related projects

HandPoseBarracuda uses the following sub-packages:

• BlazePalmBarracuda (lightweight hand detector)
• HandLandmarkBarracuda (hand/finger landmark detector)