Exploring semantic segmentation of related subclasses from a superset of classes

作者：

Highlights：

• Exploring semantic segmentation of related subclasses from a superset of classes.

• Modification of the Coco-Stuff dataset.

• Grouping of required classes to achieve higher accuracy.

• Training of model on Deeplabv3+.

• Improving visualization quality using DenseCRF.

摘要

•Exploring semantic segmentation of related subclasses from a superset of classes.•Modification of the Coco-Stuff dataset.•Grouping of required classes to achieve higher accuracy.•Training of model on Deeplabv3+.•Improving visualization quality using DenseCRF.

论文关键词：Image segmentation,Stuff classes,Deeplab

论文评审过程：Received 3 November 2020, Revised 12 November 2021, Accepted 22 December 2021, Available online 27 December 2021, Version of Record 2 January 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108509