Learning discriminative localization from weakly labeled data

作者：

Highlights：

• A framework for discriminative localization and classification is proposed.

• It works for both images and time series.

• It only requires weakly annotated data.

• It achieves results similar to those obtained with full supervision.

摘要

Highlights•A framework for discriminative localization and classification is proposed.•It works for both images and time series.•It only requires weakly annotated data.•It achieves results similar to those obtained with full supervision.

论文关键词：Discriminative discovery,Object detection,Event detection,Image classification,Time series classification,Weakly supervised learning

论文评审过程：Received 22 January 2013, Revised 14 May 2013, Accepted 24 September 2013, Available online 4 October 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.09.028