A Coding Guide to Markerless 3D Human Kinematics with Pose2Sim, RTMPose, and OpenSim

We continue the main workflow by converting the associated 2D detections into 3D coordinates via triangulation, followed by filtering to smooth out noise. Afterward, we perform marker augmentation using a trained LSTM model to predict additional virtual markers that can enhance OpenSim model scaling and inverse kinematics accuracy.

Finally, if OpenSim is available in the environment, we execute automatic model scaling and inverse kinematics computation. This step involves adjusting a generic OpenSim model to fit participant dimensions and computing 3D joint angles by fitting the model to the marker data. The output includes scaled .osim model files and joint angle .mot files.

Upon completion of these steps, we summarize the pipeline timing for each stage, providing insights into computational efficiency and performance. This summary helps in understanding the workflow's overall duration and identifying potential bottlenecks or optimization opportunities.

Here is a detailed breakdown of the final stages:

1. Triangulation: Converts 2D detections from multiple cameras into 3D coordinates.

   • Output: .trc files (OpenSim-compatible 3D marker format).

2. Filtering: Smooths out noise in the triangulated 3D trajectories using various filters like Butterworth, Kalman, etc.,

Read the full article at MarkTechPost

---

Want to create content about this topic? Use Nemati AI tools to generate articles, social posts, and more.