PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
January 24, 20263 citations

Object detection on low-compute edge SoCs: a reproducible benchmark and deployment guidelines.

View Full Paper
CKChang KongFLFeng LiXYXiaohu Yan

Key Points

  • The aim is to evaluate object detection performance on low-compute edge AI systems and identify key factors affecting efficiency.
  • Benchmarking nine YOLO variants on three Rockchip SoCs
  • Testing various input resolutions and compute configurations
  • Analyzing latency, accuracy, and energy efficiency under different conditions
  • Inference latency shows a stronger correlation with detection accuracy than FLOPs or parameter count
  • Latency scaling deviates from quadratic predictions due to bandwidth limitations
  • Memory bandwidth is crucial for performance under multitasking conditions, highlighting different efficiencies across SoCs

Abstract

Deploying deep learning-based object detectors on low-compute edge AI SoCs remains challenging, as real-world performance depends on factors beyond nominal TOPS ratings, including architectural design, memory bandwidth, and system-level contention. This study presents a comprehensive and reproducible benchmarking of nine YOLO variants across three widely used Rockchip SoCs, covering multiple input resolutions, compute configurations, and operating conditions. Our results show that inference latency correlates more strongly with detection accuracy (mAP) than with FLOPs or parameter count, revealing the execution overhead introduced by recent architectural modules. Latency scaling with input size deviates from quadratic theoretical predictions due to bandwidth limitations, and multi-core NPU scheduling provides only marginal gains because of synchronization and shared-memory bottlenecks. Under multitasking stress, memory bandwidth emerges as the primary factor governing robustness, while energy-per-inference measurements highlight substantial efficiency differences across SoCs. These findings offer practical guidance for selecting and deploying object detection models on embedded platforms, emphasizing the need for hardware-aware model choices and memory-efficient optimizations in real-time edge AI applications.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Kong et al. (2026) studied this question.

synapsesocial.com/papers/697461a8bb9d90c67120b849https://doi.org/10.1038/s41598-026-36862-y
Ask AI
Helpful
Bookmark
Share
View Full Paper