8000 GitHub - Tekicho/3dposeref: Inplementation of "3D Human Pose Refining using precise bone-lengths and 3D artificial localization errors" using PyTorch
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Inplementation of "3D Human Pose Refining using precise bone-lengths and 3D artificial localization errors" using PyTorch

Notifications You must be signed in to change notification settings

Tekicho/3dposeref

 
 

Repository files navigation

3DPoseRef: 3D Human Pose Refining using precise bone-lengths and 3D artificial localization errors

Challenges in 3D human pose estimation include occlusions (body parts being hidden), self-occlusions (body parts blocking each other), varying camera viewpoints, and the ambiguity inherent in 2D-to-3D mapping. Researchers and engineers have developed various techniques to address these challenges, such as data augmentation, multi-modal fusion, and using additional cues like clothing information.We propose a novel monocular 3D human pose refining method combining accurate bone length constraints and artificial localization errors.

Our model

Model

Demo 1: Controlling fullbody digital avatar in Unity3D

Pose1

Demo 2: Controlling fullbody digital avatar in Unity3D

Pose1

Demo 3: Controlling head pose for a digital avatar in Unity3D

Pose1

Demo 4: Controlling head pose for a digital avatar in Unity3D

Pose1

Driving avatar for qualitative evaluation

Snapshot from unity VR application Snapshot from unity VR application Snapshot from unity VR application

Training

Testing

References

MoveNet

About

Inplementation of "3D Human Pose Refining using precise bone-lengths and 3D artificial localization errors" using PyTorch

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 71.7%
  • MATLAB 18.5%
  • Lua 9.6%
  • Other 0.2%
0