SMPL-X

 Worked at the University of California, Berkeley on their project, SMPL-X. SMPL-X is a model trained by thousands of 3D scans using a neural network pose prior, which later detects gender and retrieves model parameters (shape, pose, translation) from a 2D image to fit them with the SMPL-X model and act as joints for the model.

​

I integrated SMPL-X with Unity to allow artists and developers to effortlessly express their ideas within game engines, utilizing an open-source Unity plugin. The user could then find a video on YouTube and use SMPLify to get the fitted parameters to transfer it into the plugin functions, enabling the SMPL-X model in the 3D Unity space to replicate the human movements in the video.