Unsupervised learning framework for interest point detection and description via properties optimization

作者：

Highlights：

• Unified formulation with latent variable jointly optimizes different properties.

• Informativeness property encourages features to be extracted in complicated area.

• Approximate EM algorithm efficiently maximizes the non-differentiable objective.

摘要

•Unified formulation with latent variable jointly optimizes different properties.•Informativeness property encourages features to be extracted in complicated area.•Approximate EM algorithm efficiently maximizes the non-differentiable objective.

论文关键词：Interest point,Unsupervised learning,Expectation maximization,Properties,Convolution neural network

论文评审过程：Received 6 March 2020, Revised 5 October 2020, Accepted 26 December 2020, Available online 4 January 2021, Version of Record 10 January 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107808