Automatic signboard detection and localization in densely populated developing cities

作者：

Highlights：

• Faster R-CNN based signboard localization in densely populated developing cities.

• SB_PreCNN and PreRCNN pretraining scheme for CNN backbone and RPN, respectively.

• Choosing anchor box dimensions using proposed ARAS algorithm.

• 5,200 size labeled image dataset construction based on 6 countries.

• 0.90 mAP score on test set using final model obtained after detailed evaluation.

摘要

•Faster R-CNN based signboard localization in densely populated developing cities.•SB_PreCNN and PreRCNN pretraining scheme for CNN backbone and RPN, respectively.•Choosing anchor box dimensions using proposed ARAS algorithm.•5,200 size labeled image dataset construction based on 6 countries.•0.90 mAP score on test set using final model obtained after detailed evaluation.

论文关键词：Object detection,Faster R-CNN,Clustering

论文评审过程：Received 10 April 2021, Revised 24 July 2022, Accepted 22 August 2022, Available online 27 August 2022, Version of Record 6 September 2022.

论文官网地址：https://doi.org/10.1016/j.image.2022.116857