Enabling Collaborative Video Sensing at the Edge through Convolutional Sharing

TL;DR

This paper introduces a collaborative approach for edge-based video sensing that enhances person detection accuracy by sharing scene summaries among peer nodes without retraining DNNs, reducing latency and improving recall.

Contribution

It presents a novel on-the-fly scene sharing method enabling peer collaboration in edge video sensing without retraining neural networks.

Findings

Recall improved by up to 10% with one collaborator

No re-training needed for DNNs

Minimal processing latency achieved

Abstract

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.