This paper presents a Python application to test human pose estimation (HPE) models in touchless interface research. Our software is the only solution that allows the creation and editing of arbitrary dynamic gestures, the configuration of test parameters, the tracking of body keypoints, and the generation of detailed performance metrics for comparing deep learning-based HPE models. A key motivation for developing our software was the perceived need for a unified tool to compare and evaluate AI-based human pose estimation models for touchless, gesture-based interaction across different control modalities (head-based, hand-based, and full-body) and varying user postures (standing, sitting, and moving). This software is particularly relevant in fields such as healthcare, assistive technology, rehabilitation, and augmented/virtual reality, where physical contact is impractical or undesirable. The software features a modular design that enables scalability and future development.
Dogiel et al. (Sat,) studied this question.