A deployment-centric survey published in the Journal of Cloud Computing evaluates real-time object detection within the YOLO model family across modern distributed computing environments. The authors examine YOLO not as an isolated benchmarked network, but as a scalable family of models embedded across cloud data centres, edge servers, and tiny IoT devices. Each of these deployment environments operates under distinct operational constraints, including variations in network latency, available bandwidth, onboard memory capacity, energy consumption, and infrastructure cost.
The survey begins by reviewing foundational object detection paradigms alongside the technical evolution of the YOLO architecture. To organize how these computer vision models function across heterogeneous hardware tiers, the paper introduces a macro–meso–micro perspective covering cloud, edge, and IoT deployments. The authors subsequently survey lightweight model architectures and compression methods, focusing specifically on tiny and nano model variants, efficient backbones, pruning, quantisation, and knowledge distillation. Within this framework, the review analyzes how each compression strategy influences detection accuracy, operational latency, overall model size, and energy demands.
From a systems architecture perspective, the survey categorizes deployment patterns into distinct operating models across the computing hierarchy. The authors summarize configurations including centralized cloud-centric serving, distributed edge and fog deployments, and specialized processing pipelines tailored for IoT and tiny devices. Furthermore, the analysis details multi-tier execution strategies, emphasizing collaborative hierarchical inference integrated with federated learning across network layers.
Looking toward future implementation demands, the paper outlines key open challenges and future directions for hardware-aware and deployment-aware YOLO co-design. The review underscores the importance of developing sustainable YOLO deployments, highlighting their potential contribution to United Nations Sustainable Development Goals across practical domains such as smart cities, healthcare initiatives, and environmental monitoring systems.
Subject of Research: Deployment patterns and compression techniques for YOLO object detection models across distributed systems
Article Title: Real-time YOLO across Cloud, edge, and IoT: architectures, optimisations, and deployment patterns
Article References: Attar, H., Ababneh, J., Kalaycıoğlu, A., Solyman, A., Hafez, M., Elhaty, I. A., & Solyman, A. (2026). Real-time YOLO across Cloud, edge, and IoT: architectures, optimisations, and deployment patterns. Journal of Cloud Computing. https://doi.org/10.1186/s13677-026-00976-5
Image Credits: AI Generated
DOI: 10.1186/s13677-026-00976-5
Keywords: YOLO, real-time object detection, cloud data centres, edge servers, IoT devices, distributed systems, lightweight architectures, model compression, pruning, quantisation, knowledge distillation, fog computing, hierarchical inference, federated learning, hardware-aware co-design, sustainable deployments, smart cities, environmental monitoring
News Source: Denise Maddox. (November 19, 2026). Survey Evaluates Real-Time YOLO Deployments Across Cloud, Edge, and IoT Systems. Scienmag.



