What does this research mean for the field?

A lightweight convolutional neural network architecture integrating a Ghost-Separable Block achieves superior accuracy-efficiency trade-offs for UAV onboard anomaly detection, outperforming mainstream lightweight detectors while enabling real-time processing on resource-constrained hardware. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

Develop a lightweight convolutional neural network tailored for real-time onboard visual anomaly detection and localization on resource-constrained unmanned aerial vehicle platforms.

August 27, 2026Scientific ReportsOpen Access

Lightweight convolutional neural network for UAV onboard visual anomaly detection and localization

View Full Paper

Ask AI

Bookmark

Authors

ZJZihan JiangKunming Institute of Botany GZGuiping ZhaoCapital Medical University XXXiaoqi XuShaoyang University

Discussion

Member takes

Implication

Experimental study demonstrates real-time visual anomaly detection on embedded drone processors, indicating feasible edge computing for aerial inspection.

Key Points

Develop a lightweight convolutional neural network tailored for real-time onboard visual anomaly detection and localization on resource-constrained unmanned aerial vehicle platforms.
Designed a Ghost-Separable Block combining ghost convolution, depthwise separable convolution, and coordinate attention to extract directional spatial features while reducing parameters by 26%.
Integrated a bidirectional feature pyramid network for multi-scale feature fusion and a dual-branch prediction head utilizing Complete Intersection over Union loss.
Evaluated model performance on a custom UAV aerial anomaly dataset, the public VisDrone benchmark, and an embedded Jetson Xavier NX computing platform.
Achieved 81.2% mean average precision at 0.5 IoU threshold using 2.18 million parameters, reducing parameters by 31% and computation by 59% relative to YOLOv8n.
Demonstrated real-time inference speeds on a Jetson Xavier NX of 42 frames per second natively and 67 frames per second with TensorRT FP16 within a 15 W power budget.

Cite This Study

Jiang et al. (2026) studied this question.

synapsesocial.com/papers/6a8fe98110c91c1e926211e1 https://doi.org/10.1038/s41598-026-68007-6

View Full Paper

Ask AI

Bookmark

Implication

Experimental study demonstrates real-time visual anomaly detection on embedded drone processors, indicating feasible edge computing for aerial inspection.

Lightweight convolutional neural network for UAV onboard visual anomaly detection and localization

Discussion

Member takes

Implication

Key Points

Cite This Study

Lightweight convolutional neural network for UAV onboard visual anomaly detection and localization

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider