Screw defect detection using Xception algorithm over Resnet50 algorithm for better accuracy

View at publisher Summarise with VIRAT
Authors: Tirumalasetty Nithin Babu, T. Yaswanth Kumar
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Aim: The primary aim of our research is to enhance the precision of screw defect identification through the use of sophisticated machine learning models, focusing specifically on the comparison between Xception and Resnet. The objective isto assess and boost the prediction accuracyof errors. In this investigation, Xception demonstrated an accuracy of 92%, whereas Resnet reached 90% accuracy. The sample size for every model remained constant at 10. Xception’s mean error prediction was 81.00 with a standard deviation of 6.68, while Resnet’s mean error was 78. Materials and Methods: This study used two different deep learning architectures, Xception and Resnet, to optimize screw defect detection. The dataset was carefully split into two sets based on each model to ensure a consistent sample size of 10 for Xception and Resnet. To measure the performance of each model, a comprehensive statistical analysis was performed using appropriate tools and assuming a significance level of 0.05. Results : The results show that Xception outperformed Resnet in detecting screw defects, achieving 92% accuracy, while Resnet achieved his 90% accuracy. Statistical evaluation showed a notable difference between the two models, with a p-value of 0.0085 (p

Fields of Study & Subjects

Industrial Vision Systems and Defect Detection Advanced Neural Network Applications Non-Destructive Testing Techniques

Publication details

Journal / source
CRC Press eBooks
Pages
51–58
Publication date
25 March 2026
Language
EN
Document type
Book chapter