PoseLandmarkBarracuda

This is pose landmark detector that works with a monocular color camera.
This works by using the MediaPipe Pose model in Unity Barracuda.
It also includes a simple segmentation function.

PoseLandmarkBaracuda is based on HandLandmarkBarracuda implementation by keijiro.

ONNX Model

The ONNX model files used here have been converted for Unity Barracuda using tf2onnx --opset=9, and the same kind of operation as keijiro's BlazeFaceBarracuda. See his Colab notebook for details.

Require

• Unity 2020

Install

It can be installed by adding scoped registry to the manifest file(Packages/manifest.json).

scopedRegistries

{
    "name": "torinos",
    "url": "https://registry.npmjs.com",
    "scopes": ["jp.torinos"]
}

dependencies

"jp.torinos.mediapipe.poselandmark": "1.0.0"