Ship target detection algorithm based on lightweight YOLOv7-tiny
To solve the problem of the large number of param-eters and computation of ship target detection algorithm,as well as the difficulties of ship detection caused by the influ-ence of the nearshore complex backgrounds and the mutual oc-clusion of ships in inland river environments,a lightweight al-gorithm termed MED-YOLO for ship target detection was pro-posed by virtue of improvements on YOLOv7-tiny.Firstly,the MobileNetV3 network was used as the backbone feature ex-traction network,thereby greatly reducing the calculation cost of the model.Secondly,the EMA attention module was intro-duced into the neck network,thereby contributing to the EMA-ELAN module that enhanced multi-dimensional percep-tion and multi-scale feature extraction capability of the net-work.Then,combining with scale,spatial,and task percep-tions,the Dyhead was selected as the detection head of the improved model to obtain stronger feature expression ability.Finally,the WIoU with dynamic non-monotonic focusing mechanism was used as the bounding-box loss function to cope with ship occlusion and improve the detection performance.The experimental results show that,compared with YOLOv7-tiny,the proposed MED-YOLO has 39.8％fewer parameters and 55.0％less computation,and its precision and mAP@0.5 have increased by 1.4％and 1.0％,respectively,reaching 98.3％and 98.9％,which not only achieves lightweight,but also has better detection performance,which meets the de-ployment requirements in limited computing resources,and thereby offering practical engineering significance.
YOLOv7-tinyobject detectionlightweightat-tention mechanismloss function
NETL
NSTL
万方数据
