基于CBAM和BiFPN改进YoloV5的渔船目标检测-豆柴文库

您所在位置：网站首页 / 基于CBAM和BiFPN改进YoloV5的渔船目标检测.docx / 文档详情

在线预览结束，喜欢就下载吧，查找使用更方便

5 金币

下载文档

/ 2

下载提示

如果您无法下载资料，请参考说明：

1、部分资料下载需要金币，请确保您的账户上有足够的金币

2、已购买过的文档，再次下载不重复扣费

3、资料包下载后请先用软件解压，在使用对应软件打开

文本预览

基于CBAM和BiFPN改进YoloV5的渔船目标检测
Title:EnhancedYOLOV5-basedFishermanVesselDetectionwithCBAMandBiFPN
Abstract:
Fishermanvesseldetectionplaysacrucialroleinmaritimesafety,environmentalconservation,andfisherymanagement.However,itremainsachallengingtaskduetothecomplexnatureofthemarineenvironment.Inthispaper,weproposeanenhancedYOLOV5modelforfishermanvesseldetectionbyincorporatingtheChannelAttentionModule(CBAM)andBi-directionalFeaturePyramidNetwork(BiFPN).Ourmethodaimstoimprovetheaccuracy,robustness,andefficiencyoffishermanvesseldetection,leadingtobettermaritimesurveillanceandresourcemanagement.
1.Introduction
Fishermenvesselsareessentialfortheglobalfishingindustry,providinglivelihoodsformillionsofpeople.However,theincreasingnumberoffishingactivitiesandthepotentialforoverfishinghaveraisedseriousconcernsaboutsustainabilityandfisherymanagement.Accurateandefficientfishermanvesseldetectionisessentialformonitoringandregulatingfishingactivities,ensuringtheprotectionofmarineecosystems.ThispaperpresentsanenhancedYOLOV5modelforimprovedfishermanvesseldetectionusingCBAMandBiFPN.
2.RelatedWork
Thefieldofobjectdetectionhasseensignificantadvancementsinrecentyears,withvariousmodelsproposed,suchasYOLO,FasterR-CNN,SSD,andEfficientDet.Manyofthesemodelshavebeenappliedtofishermanvesseldetectionwithvaryingdegreesofsuccess.TheCBAMmodulefocusesonthechannelandspatialattentionmaps,enablingthemodeltoselectivelyamplifyinformativefeatures.BiFPNimprovesfeaturerepresentationbyfusingfeaturesfrommultiplelevelsofthefeaturepyramid.
3.Methodology
3.1YOLOV5
WechooseYOLOV5asourbasemodelduetoitsexcellentbalancebetweenaccuracyandefficiency.YOLOV5consistsofabackbonenetworkforfeatureextractionandapredictionheadforobjectlocalizationandclassification.
3.2CBAMIntegration
WeintroducetheCBAMmoduleintothebackbonenetworkofYOLOV5toenhancefeaturerepresentation.Byincorporatingchannelandspatialattentionmaps,themodelcanfocusoninformativefeatureswhilesuppressingirrelevantones.Thishelpsthemodeltolearndiscriminativefeaturesforfisher