Visual tracking via weakly supervised learning from multiple imperfect oracles

作者：

Highlights：

• A novel weakly supervised learning-based object tracking method is proposed.

• The method presents a natural way of fusing multiple complementary methods.

• The method can evaluate the online tracking methods in the absence of ground truth.

摘要

•A novel weakly supervised learning-based object tracking method is proposed.•The method presents a natural way of fusing multiple complementary methods.•The method can evaluate the online tracking methods in the absence of ground truth.

论文关键词：Visual tracking,Weakly supervised learning,Information fusion,Online learning,Adaptive appearance model,Drift problem,Online evaluation

论文评审过程：Received 6 June 2012, Revised 28 April 2013, Accepted 2 October 2013, Available online 11 October 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.10.002