MonoPoly: A practical monocular 3D object detector

作者:

Highlights:

• Efficient Monocular 3D Detector Splitting Configuration with Accuracy-Speed Trade-off.

• Contextual projection for auxiliary learning and inference free gains.

• Multi-scale feature seamless aggregation and upsampling by PolyPAN module.

• Attention-aware loss for regression weight guidance.

摘要

•Efficient Monocular 3D Detector Splitting Configuration with Accuracy-Speed Trade-off.•Contextual projection for auxiliary learning and inference free gains.•Multi-scale feature seamless aggregation and upsampling by PolyPAN module.•Attention-aware loss for regression weight guidance.

论文关键词:Object detection,Monocular 3D,Real-time,Light-weight

论文评审过程:Received 14 April 2022, Revised 19 July 2022, Accepted 9 August 2022, Available online 10 August 2022, Version of Record 20 August 2022.

论文官网地址:https://doi.org/10.1016/j.patcog.2022.108967