Journal of Engineering, Project, and Production Management, 2026, 16(3), 2026-226
AFB-YOLO: An Improved Infrared Ship Detection Method for Maritime Applications Based on YOLOv7
Undergraduate Student,
Queen Mary Hainan College, Beijing University of Posts and
Telecommunications
Lingshui, Hainan, 572400, China, E-mail: jp2023213791@qmul.ac.uk
Project Management
Received February 10, 2026; revised April 23, 2026; accepted April 26, 2026
Available online May 4, 2026
Abstract: Infrared ship images suffer from issues such as blurred target features, complex backgrounds, and a high proportion of small targets, which render existing detection methods prone to missed detections and false alarms. This paper proposes the Attention and Feature Balanced You Only Look Once (AFB-YOLO) method. The model adopts YOLOv7 (You Only Look Once version 7) as its baseline architecture. In the backbone network, an ELAN-O module is introduced to enhance feature extraction capability, ODConv dynamic convolution is employed to adaptively perceive multi-scale features, and a Similarity-based Attention Module (SimAM) is incorporated to intensify focus on ship regions. Within the neck structure, a weighted fusion mechanism is constructed to improve the effect of path aggregation. The detection head utilizes an Adaptively Spatial Feature Fusion (ASFF) module to mitigate spatial feature conflicts and elevate multi-scale perception. With respect to the loss function, the Efficient Intersection over Union (EIOU) loss is introduced to optimize the bounding box regression process and improve localization accuracy. Experimental evaluations on a public infrared ship dataset demonstrate that AFB-YOLO achieves an mAP@0.5 (mAP@0.5 stands for mean Average Precision at Intersection over Union (IoU) threshold 0.5) of 90.7%, markedly outperforming the original YOLOv7 and surpassing YOLOv12 by 12.3 percentage points. These results confirm that the proposed method effectively addresses the inherent limitations of infrared imaging and substantially enhances ship detection performance.
Keywords: Ship detection; feature extraction; dynamic convolution; attention mechanism; path Aggregation.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Li, J. (2026). AFB-YOLO: An Improved Infrared Ship Detection Method for Maritime Applications Based on YOLOv7. Journal of Engineering, Project, and Production Management, 16(3), 2026-226.
