Rex Tsou

ICASI Conference Paper Abstract

We propose an innovative generative AI model that significantly enhances the accuracy of human pose estimation, a critical component in the realm of motion detection and activity tracking. The model ingeniously interprets absent skeletal points as distortions in the visual data, akin to noise. By employing sophisticated denoising algorithms, it can precisely infer and restore these occluded or unobserved skeletal points. Such an approach substantially refines the precision of pose predictions, which is immensely beneficial across various demanding applications. For instance, in crowded environments, it ensures the distinct identification of individual poses amidst the chaos. Lastly, the model's application in virtual reality results in highly immersive and interactive user experiences due to its ability to replicate the nuanced movements of the human body with remarkable fidelity. By pushing the boundaries of what's possible with human pose estimation, this generative AI model unlocks new levels of reliability and functionality in human body pose analysis.