Effect of Compressed Sensing Rates and Video Resolutions on a PoseNet Model in an AIoT System

Abstract

To provide an artificial intelligence service such as pose estimation with a PoseNet model in an Artificial Intelligence of Things (AIoT) system, an Internet of Things (IoT) sensing device sends a large amount of data such as images or videos to an AIoT edge server. This causes serious data traffic problems in IoT networks. To mitigate these problems, we can apply compressed sensing (CS) to the IoT sensing device. However, the AIoT edge server may have poor pose estimation accuracy (i.e., pose score), because it has to recover the CS data received from the IoT sensing device and estimate human pose from the imperfectly recovered data according to CS rates. Therefore, in this paper, we analyze the effect of CS rates (from 100% to 10%) and video resolutions (1280×720, 640×480, 480×360) in the IoT sensing device on the pose score of the PoseNet model in the AIoT edge server. When only considering the meaningful range of CS rates from 100% to 50%, we found that the higher the video resolution, the lower the pose score. At the CS rate of 80%, we could reduce data traffic by 20% despite the degradation in pose score of less than about 0.03 for all video resolutions.

Detection of 3D Human Posture Based on Improved Mediapipe

“…PoseNet [9] is a technical solution for posture estimation released by Tensor-Flow and Google Creative Lab. PoseNet can also be used to estimate a single pose or multiple poses, and it is not too dependent on the performance acceleration of GPU, but it is also unable to recognize 3D human posture data.…”

Connecting artificial intelligence to value creation in services: mechanism and implications

“…Cloud platform (Topic 2) addresses the use of sensors and cloud servers for data collection. In particular, this topic covers sensing technologies necessary for supporting efficient data collection (Kwon and Seo 2022) and frameworks for virtual sensor configurations to collect data via multiple devices (Alberternst et al 2021). Learning technology (Topic 3) addresses the machine learning algorithms can be used in various AI service areas (e.g., smart city service and healthcare service), such as deep learning models for image classification and segmentation (Lee et al 2022b;Tseng et al 2021) and federated learning models for privacy protection of AI services (Rodríguez-Barroso et al 2020).…”

Current Research in Future Information and Communication Engineering 2022

“…The first paper, authored by Shin et al [12], proposed and evaluated an intelligent monitoring framework based on the IoT by applying a recurrent neural network (RNN) for the predictive maintenance of a biobanking system in real time. The second paper, authored by Kwon et al [13], analyzed the effect of compressed sensing (CS) rates (from 100% to 10%) and video resolutions (1280 × 720, 640 × 480, 480 × 360) in the IoT sensing device on the pose score of the PoseNet model in the artificial intelligence of things (AIoT) edge server. In this article, when the CS rate is 80% and the video resolution is 1280 × 720, the most effective data traffic mitigation and pose score were achieved.…”