DaSiamRPN: Zheng Zhu, Qiang Wang, Bo Li, Wu Wei, Junjie Yan, Weiming Hu."Distractor-aware Siamese Networks for Visual Object Tracking." ECCV (2018).
文章主要贡献:1.作者觉得SiamRPN与SiamiseFC的训练集中样本不均衡
(1).加入Detection pair (ImageNet,COCO中做数据增广)
(2).negative simple in same categories (Called Distractor-aware Training)
(3).negative simple in different categories (Called Distractor-aware Training)
2.引入Distractor Model,将Proposal与exemplar的相似性度量得分减去所有之前预先得到的Distrator(NMS将网络 提 出的proposal去冗余,去掉classification score最高的proposal,在剩下的Distrator set 中保留score大于给定阈 值的proposal)与当前proposal 的score(相似性度量)的加权和的平均
3. long-term tracking(当追丢时用local to global stategy 恒定step迭代的增加搜索区域的大小)
图里