基于改进GoogLeNet的试卷分数识别算法

打开文本图片集

中图分类号：TP391.1 文献标识码：A 文章编号：2096-4706（2025）18-0053-06

Abstract： Curently，there are problemsof time-consuming，cumbersomeness，high repetitiveness and low efficiency whenteachers inputcollegestudentstestpaperscores.Toefectivelyimprovetheaccuracyofdigitalrecognition intestpapers， this studyconstructsatestscorerecognitionalgorithmI-GoogLeNetbasedonimproved GoogLeNet.Firstly，itadjusts the contrastthroughthehistogramequalizationalgorithm CLAHEtohighlight thehandwritingof theredscores.Secondly，itadds anauxiliaryclasifertotheGoogLeNetarchitectureandreplaces theactivationfunctionwith Swishtorecognizetestcores. Finaly，itusestheApacheOibraytoexportandsvethesoresExperimentssowatoartifcialdatasets，teecoition accuracy of I-GoogLeNet is higher than that of GoogLeNet andLeNet-5 by 3.75% and 6.14% ，respectively.

Keywords： test score recognition; CLAHE; GoogLeNet; auxiliary classifier; Convolutional Neural Networks

0 引言

在教育领域，试卷成绩是检验学习成果的重要依据。（剩余9330字）