Real-Time High-Resolution Pedestrian Detection in Crowded Scenes via Parallel Edge Offloading

TL;DR

Hode is a framework that accelerates high-resolution pedestrian detection in crowded scenes by intelligently offloading image regions to multiple edge nodes, balancing speed and accuracy.

Contribution

It introduces a novel parallel offloading approach with context-aware region partitioning and DRL-based scheduling for real-time high-res pedestrian detection.

Findings

Achieves up to 2.01% speedup with minimal accuracy loss

Utilizes spatio-temporal flow filtering for effective region partitioning

Employs DRL-based scheduling for load balancing among edge nodes

Abstract

To identify dense and small-size pedestrians in surveillance systems, high-resolution cameras are widely deployed, where high-resolution images are captured and delivered to off-the-shelf pedestrian detection models. However, given the highly computation-intensive workload brought by the high resolution, the resource-constrained cameras fail to afford accurate inference in real time. To address that, we propose Hode, an offloaded video analytic framework that utilizes multiple edge nodes in proximity to expedite pedestrian detection with high-resolution inputs. Specifically, Hode can intelligently split high-resolution images into respective regions and then offload them to distributed edge nodes to perform pedestrian detection in parallel. A spatio-temporal flow filtering method is designed to enable context-aware region partitioning, as well as a DRL-based scheduling algorithm to allow accuracy-aware load balance among heterogeneous edge nodes. Extensive evaluation results using realistic prototypes show that Hode can achieve up to 2.01% speedup with very mild accuracy loss.