Exploiting spatial relation for fine-grained image classification

作者：

Highlights：

• Spatial relation between object parts is considered to construct discriminative representation of fine-grained image for FGIC for the first time.

• A selection strategy is proposed to find the useful spatial relations. The finegained image can be represented by exploiting the interaction between object parts.

• The proposed method is evaluated on CUB-200-2011 and FGVC Aircraft datasets and the experiment results prove the effectiveness of the proposed method.

摘要

•Spatial relation between object parts is considered to construct discriminative representation of fine-grained image for FGIC for the first time.•A selection strategy is proposed to find the useful spatial relations. The finegained image can be represented by exploiting the interaction between object parts.•The proposed method is evaluated on CUB-200-2011 and FGVC Aircraft datasets and the experiment results prove the effectiveness of the proposed method.

论文关键词：Fine-grained image classification,Spatial relation,Convolutional neural network

论文评审过程：Received 15 May 2018, Revised 6 January 2019, Accepted 10 February 2019, Available online 15 February 2019, Version of Record 20 February 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.02.007