LifelongGlue: Keypoint matching for 3D reconstruction with continual neural networks

作者：

Highlights：

• Continual learning with knowledge consolidation, retention and reusability.

• Exploitation of continual graph attention network for image keypoint matching.

• Model capable of estimating accurate keypoint matches.

• Performance analysis for Camera pose estimation in 3D reconstruction.

• Better performance accuracy as compared to other neural network methods.

摘要

•Continual learning with knowledge consolidation, retention and reusability.•Exploitation of continual graph attention network for image keypoint matching.•Model capable of estimating accurate keypoint matches.•Performance analysis for Camera pose estimation in 3D reconstruction.•Better performance accuracy as compared to other neural network methods.

论文关键词：Continual learning,Graph neural networks,Attention networks,Image keypoint matching,3D scene reconstruction

论文评审过程：Received 26 October 2021, Revised 7 January 2022, Accepted 22 January 2022, Available online 5 February 2022, Version of Record 8 February 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116613