Scale-balanced loss for object detection

作者：

Highlights：

• A matching imbalance in current object detection pipelines is pointed out. It can lead to poor performance of detecting objects with different scales.

• An innovative loss function called scale-balanced loss is proposed to alleviate the matching imbalance.

• Experiments demonstrate the effectiveness of the scale-balanced loss, especially the performance of detecting small objects is improved significantly.

摘要

•A matching imbalance in current object detection pipelines is pointed out. It can lead to poor performance of detecting objects with different scales.•An innovative loss function called scale-balanced loss is proposed to alleviate the matching imbalance.•Experiments demonstrate the effectiveness of the scale-balanced loss, especially the performance of detecting small objects is improved significantly.

论文关键词：Object detection,Neural network,Matching imbalance

论文评审过程：Received 6 November 2019, Revised 15 March 2021, Accepted 18 April 2021, Available online 24 April 2021, Version of Record 8 May 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107997