Robust object representation by boosting-like deep learning architecture

作者：

Highlights：

• The proposed method is on the combination of the SMF structure and a boosting-like algorithm.

• The SMF structure is based on the fusion of the handcraft features and deep learned features.

• We obtain more convergence stability than traditional deep learning algorithm.

摘要

•The proposed method is on the combination of the SMF structure and a boosting-like algorithm.•The SMF structure is based on the fusion of the handcraft features and deep learned features.•We obtain more convergence stability than traditional deep learning algorithm.

论文关键词：Boosting,Deep learning,Object representation,Synchronized feature

论文评审过程：Received 10 December 2015, Revised 11 May 2016, Accepted 2 June 2016, Available online 3 June 2016, Version of Record 29 September 2016.

论文官网地址：https://doi.org/10.1016/j.image.2016.06.002