Scalable edge cloud system improves multi-site video anomaly detection
YUTELA: A Scalable Edge Cloud AIoT System for Multi-Site Video Anomaly Detection with Scene Specific Adaptation
Computer Vision and Pattern Recognition
Summary
Video systems that detect strange events often rely on powerful computers on-site. The authors developed YUTELA, which uses smaller edge devices and cloud servers to watch many cameras at once and adjust detection to each location. This setup runs faster than older systems and can handle different activity levels at multiple public sites. It retrains its models automatically using recent data and keeps alerts timely with low delay. This makes it practical to monitor many places efficiently and adapt to changing conditions.
What this means in practice
- •For security system integrators: Deploy multi-site video anomaly detection with scalable edge and cloud processing for improved real-time monitoring and scene-specific model adaptation.$Commercial implications: Enables selling advanced video surveillance products with edge-cloud AI that adapts to each camera’s environment for security companies.
- •For smart city operations teams: Manage and monitor multiple public cameras efficiently by utilizing edge devices and centralized retraining to adapt anomaly detection to each site’s activity patterns.
Authors
Shanle Yao, Babak Rahimi Ardabili, Amir Kimiyaie, Hamed Tabkhi
Abstract
Building upon our previous Ancilia framework, this work presents \emph{YUTELA}, an edge--server--cloud AIoT system for long-term, multi-site video anomaly detection. YUTELA replaces workstation-based on-site processing with NVIDIA Jetson Thor edge devices and redesigns the multi-camera pipeline through camera batching and redistribution, shared memory, reduced inter-stage data transfer, task integration, and TensorRT-accelerated inference. It maintains a separate anomaly model per camera, retrained on a centralized GPU server once enough recent pseudo-normal data are collected, with updated weights returned to the edge. With eight cameras, TensorRT increases throughput from 14.17 to 29.94 FPS under normal density, and the optimized edge node exceeds the throughput of the workstation-based Ancilia system in seven of nine workload settings, reaching up to 2.03 times its throughput at extreme density. This gain comes at the cost of higher latency under lighter loads with four or eight streams. YUTELA was then operated continuously for 170 hours at four public sites using 31 existing CCTV cameras. The deployment shows that runtime load is driven by scene activity and input frame rate rather than by camera count, and that retraining frequency varies by about three times across sites. The 101 automatically triggered training events occupied only about 6.6\% of a single GPU, indicating that one global node can support adaptation for tens of sites. Mean end-to-end alert latency ranged from 6.8 to 15.4~s depending on site traffic and anomaly type. These results show that YUTELA enables practical multi-camera edge processing with low-cost, camera-specific adaptation across heterogeneous sites.