Human-Motion-Reconstruction

This project combines EasyMocap's Github project and Python's Django software to develope a user interface that automated the process of multi-video syncrhonization, intrinsic and extrinsic camera calibration, and human motion capture and reconstruction.

---

Process

Input video:

You can watch the input video here.

YOLO is used for human detection and HRNet is used for 2D keypoint detection:

Multi-View Triangulation is used to estimate accurate 3D keypoints:

The SMPL model is fit to the 3D keypoints to obtain an occurate mesh:

vis_render.mp4

---

User Interface Demo

You can watch the demonstration video here

Mac Download Instructions

Check out the Mac Instructions PDF to view download instructions