Optical-flow-based framework to boost video object detection performance with object enhancement

作者：

Highlights：

• A new video object detection framework with object enhanced to solve the object occlusion problem.

• Combination of lightweight optical flow network and object detection network.

• Background information masking method boosts object detection performance.

• Performance of the proposed framework is evaluated on ImageNet VID dataset.

摘要

•A new video object detection framework with object enhanced to solve the object occlusion problem.•Combination of lightweight optical flow network and object detection network.•Background information masking method boosts object detection performance.•Performance of the proposed framework is evaluated on ImageNet VID dataset.

论文关键词：Video Object Detection,Optical Flow Network,Feature Fusion,ImageNet VID

论文评审过程：Received 28 September 2019, Revised 19 December 2020, Accepted 24 December 2020, Available online 30 December 2020, Version of Record 7 January 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.114544