SCL—Segmentation–Classification combined Loss for surface defect detection

作者：

Highlights：

• New multi-task combined loss for surface defect detection tasks.

• New Oliena dataset for the poorly researched task of glass bottles defect detection.

• Comparison between classification and segmentation networks for defect detection.

• Comparison between a multi-task loss function and a multi-branch architecture.

摘要

•New multi-task combined loss for surface defect detection tasks.•New Oliena dataset for the poorly researched task of glass bottles defect detection.•Comparison between classification and segmentation networks for defect detection.•Comparison between a multi-task loss function and a multi-branch architecture.

论文关键词：Deep learning,Segmentation model,Combined loss function,Glass bottles,Surface defect detection,Dataset

论文评审过程：Received 21 May 2021, Revised 6 February 2022, Accepted 20 February 2022, Available online 7 March 2022, Version of Record 27 March 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116710