Feedback-driven loss function for small object detection

作者:

Highlights:

• A novel Feedback-driven loss function for object detection is proposed in this paper which can train the detectors in a more balanced way.

• The loss proportion information is introduced into the loss calculation as the feedback signal.

• The detectors trained by the Feedback-driven loss function achieve higher performance on the small object detection.

• The Feedback-driven loss function is easy to integrate into the major object detectors.

摘要

•A novel Feedback-driven loss function for object detection is proposed in this paper which can train the detectors in a more balanced way.•The loss proportion information is introduced into the loss calculation as the feedback signal.•The detectors trained by the Feedback-driven loss function achieve higher performance on the small object detection.•The Feedback-driven loss function is easy to integrate into the major object detectors.

论文关键词:Small object detection,Loss function,Feedback-driven,Loss distribution balance

论文评审过程:Received 21 April 2021, Accepted 2 May 2021, Available online 7 May 2021, Version of Record 18 May 2021.

论文官网地址:https://doi.org/10.1016/j.imavis.2021.104197