科技创新与应用2025，Vol.15Issue(13)：14-19,23,7.DOI:10.19981/j.CN23-1581/G3.2025.13.004
基于深度学习的车辆部件半自动标注研究
邵延富 1谢大为1
作者信息
- 1. 广州市景心科技股份有限公司,广州 510000
- 折叠
摘要
Abstract
In the era of artificial intelligence,intelligent loss assessment technology using deep learning is a general trend in the development and transformation of automobile insurance.It can greatly reduce the cost of manual loss assessment and improve the efficiency of loss assessment;deep learning models require a large number of annotated samples for training,but currently,there is no public data set on vehicle injury types and components.This paper proposes a semi-automatic labeling method for vehicle parts,first,a model that can be roughly labeled is trained through a seed dataset,and then the model is used for automatic labeling of vehicle parts;on the basis of automatic labeling,a small number of manual corrections are made,and then expanded to the dataset,and the model is re-trained to improve the accuracy of re-labeling.The experimental results show that the semi-automatic labeling method can effectively improve the labeling efficiency of vehicle parts and reduce the labor and time costs required for data labeling work.关键词
深度学习/图像分割/半自动标注/车辆部件/数据Key words
deep learning/image segmentation/semi-automatic annotation/vehicle parts/data分类
信息技术与安全科学引用本文复制引用
邵延富,谢大为..基于深度学习的车辆部件半自动标注研究[J].科技创新与应用,2025,15(13):14-19,23,7.
