SEMEDA: Enhancing segmentation precision with semantic edge aware loss

作者：

Highlights：

• We highlight several pitfalls of the traditional per-pixel cross-entropy loss for semantic segmentation.

• We propose a novel SEMEDA loss function that uses a Semantic Edge Detection Network to avoid these pitfalls.

• Experiments on several datasets shows that SEMEDA improves the segmentation accuracy with negligible computational overhead.

摘要

•We highlight several pitfalls of the traditional per-pixel cross-entropy loss for semantic segmentation.•We propose a novel SEMEDA loss function that uses a Semantic Edge Detection Network to avoid these pitfalls.•Experiments on several datasets shows that SEMEDA improves the segmentation accuracy with negligible computational overhead.

论文关键词：Semantic segmentation,Loss function,Computer vision

论文评审过程：Received 14 January 2020, Revised 21 May 2020, Accepted 20 July 2020, Available online 31 July 2020, Version of Record 5 August 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107557