Enabling Collaborative Video Sensing at the Edge through Convolutional Sharing
This work aims to improve the accuracy of person detection for AIoT-based sensing applications by enabling collaborative processing among edge devices, which is an incremental improvement for this domain.
This paper addresses the challenge of deploying deep neural networks (DNNs) in AIoT-based sensing applications due to their high computational complexity. The authors propose a collaborative video sensing paradigm where peer nodes share scene summaries to improve person detection accuracy, achieving up to a 10% increase in recall with a single collaborator without retraining DNNs.
While Deep Neural Network (DNN) models have provided remarkable advances in machine vision capabilities, their high computational complexity and model sizes present a formidable roadblock to deployment in AIoT-based sensing applications. In this paper, we propose a novel paradigm by which peer nodes in a network can collaborate to improve their accuracy on person detection, an exemplar machine vision task. The proposed methodology requires no re-training of the DNNs and incurs minimal processing latency as it extracts scene summaries from the collaborators and injects back into DNNs of the reference cameras, on-the-fly. Early results show promise with improvements in recall as high as 10% with a single collaborator, on benchmark datasets.