DSP: Discriminative Spatial Part modeling for Fine-Grained Visual Categorization

作者：

Highlights：

• Only with the bounding-box annotations, we propose the Discriminative Spatial Part model for Fine-Grained Visual Categorization.

• A bounding box transfer learning is proposed for object localization.

• With the primary orientation of the object, the DSP could generate and select the discriminative parts to describe the fine-grained object.

摘要

•Only with the bounding-box annotations, we propose the Discriminative Spatial Part model for Fine-Grained Visual Categorization.•A bounding box transfer learning is proposed for object localization.•With the primary orientation of the object, the DSP could generate and select the discriminative parts to describe the fine-grained object.

论文关键词：Orientational Spatial Part model,Discriminative Spatial Part modeling,Fine-Grained Visual Categorization,CNN

论文评审过程：Received 29 July 2016, Revised 10 January 2017, Accepted 13 May 2017, Available online 22 May 2017, Version of Record 29 May 2017.

论文官网地址：https://doi.org/10.1016/j.imavis.2017.05.003