Edge Computing of Human Poselet

T. Byrwa, J. Kłopotek Główczewski, M. Czubenko (Gdansk Univ. of Techn., Poland)

The study investigated the computational capabilities of the NVIDIA Jetson Orin Nano -- one of the most popular edge computing platforms -- in the domain of Human Pose Estimation (HPE). A comparative evaluation was conducted across several widely adopted pose estimation architectures, assessing performance in GPU execution mode. Key performance indicators included inference latency, power consumption, memory footprint, and processor utilization. The goal was to determine the feasibility of deploying real-time pose estimation models on resource-constrained edge devices. Experimental results highlight the trade-offs between model complexity and system efficiency, offering practical insights for the edge-based deployment of deep learning models.