This proyect is a system body tracking using Machine Learning in Unity through Barracuda to move avatar 3d model
This proyect is a system body tracking using Machine Learning in Unity through Barracuda, using a training model (https://digital-standard.com/threedpose/models/Resnet34_3inputs_448x448_20200609.onnx) in ONX to recognice the body from a video or webcam to transfer the joint position to a model 3d.
This project is based on the "ThreeDPoseUnityBarracuda”, developed by Hinori (https://github.com/digital-standard/ThreeDPoseUnityBarracuda).
This is the new tutorial with unity package to Unity: https://youtu.be/p-ZyqOd9Wqs
This is the steps for Unity 2019 In this project you can add easly a 3d model, following the next steps:
- Download file .onnx model: https://drive.google.com/file/d/1yviZ3Mze8MNf2TYpYA12_p66YOzfkmvY/view?usp=sharing
- Copy the file .onnx on folder: BarracudaBodyTracking\Assets\Scripts\Model
- Open SampleScene from Scenes folder
- Put the .onnx file into NNModel checkbox, ininspector of BarracudaRunner Component of the herarchy (https://drive.google.com/drive/folders/1CG-R0I8QfW8EGsgMyWn7Yk1pwOeK2DR9?usp=share_link)
- Unity -> Play
If you like to change model 3d Avatar you need:
- Add 3d model with rigging (bones for al the body include hands finger bones) in Unity project.
- Modify the features of the rigging, using the tool Rig/Configure of Unity.
- The bones are automatically associated with the components of the 3d model.
- Put the 3d model on the scene and add the script VNectModel to the 3d model.
- On BarracudaRunner Object in inspector select the new 3d model on VNectModel section.