基于深度学习的车辆部件半自动标注研究

打开文本图片集

中图分类号：TP391.4 文献标志码：A 文章编号：2095-2945（2025）13-0014-07

Abstract：Intheraofartifialinteligene，intelligentlossassssmenttechnologyusingdeplearningisageneraltrendin thedevelopmentandtransformationofautomobileinsurance.Itcangreatlyreduce thecostofmanuallossassessmentand improvetheefciencyoflossassessment;deeplearningmodelsrequirealargenumberofannotatedsamplesfortraining，but currentlythereisnopublicdatasetonvehicleinjurytypesandcomponents.Thispaperproposesasemi-automaticlabeling methodforvehicleparts，first，amodelthatcanberoughlylabeledistrainedthroughaseddataset，andthenthemodelis usedforautomaticlabelingofvehicleparts;onthebasisofautomaticlabeling，asmallnumberofmanualcorectionsaremade， andthenexpandedtothedataset，andthemodelisre-trainedtoimprovetheaccuracyofre-labeling.Theexperimentalresults showthatthesemi-automaticlabelingmethodcanefectivelyimprovethelabelingefciencyofvehiclepartsandreducethe labor and time costs required for data labeling work.

Keywords： deep learning; image segmentation; semi-automatic annotation; vehicle parts; data

据公安部统计，截至2024年底，我国汽车保有量突破4.5亿辆。（剩余9277字）