TRAC3D: Tracking and Recognition of Actions in Communication 3D
https://github.com/naotoienaga/TRAC3D
TRAC3D is an open-source system for the three-dimensional visualisation and quantitative analysis of human movement in communicative contexts. Developed by Naoto Ienaga and Kazuki Sekine, this platform enables researchers to reconstruct 3D motion trajectories from ordinary 2D video recordings. By combining camera calibration and pose-estimation algorithms, TRAC3D provides a simple workflow for examining gestures, postures and action spaces in three dimensions — all without the need for dedicated motion-capture hardware.
The system runs entirely on Google Colab, requiring no local installation. Two interactive notebooks guide users through the full process: camera calibration, 3D pose estimation, and motion visualisation. The resulting 3D plots allow users to explore spatial gesture patterns, wrist trajectories, and movement heatmaps from any perspective. TRAC3D has been presented at the International Society for Gesture Studies (ISGS 2025) and other conferences, contributing to the advancement of multimodal communication research through accessible computational tools.
