Siamese network: 连接两个输入并产生一个输出。目的是确定输入到网络的两个图像补丁中是否存在相同的物体。网络测量两个输入之间的相似度，并具有共同学习相似度和特征的能力。
Siamese+CF:
SiamFC 《Fully-convolutional siamese networks for object tracking ECCV, 2016》 GPU 速度，3 尺度 86fps，5 尺度 58fps
SiameseRPN CVPR 2018 160fps
18年ECCV的 DaSiameseRPN 160fps
改进一 CFNet 《End-to-end representation learning for Correlation Filter based tracking CVPR2017》GPU 75fps
改进二 DCFNet 《Discriminant Correlation Filters Network for Visual Tracking ICIP, 2017》100FPS
SiamRPN++：Evolution of Siamese Visual Tracking with Very Deep Networks 35fps
SiamFC++：Towards Robust and Accurate Visual Tracking with Target Estimation 90FPS
NoN-Correlation Filter Trackers
Patch learning Tracker:
MDNet 《Learning multi-domain convolutional neural networks for visual tracking CVPR, 2016》 2015年VOT冠军
改进：《SANet: Structure-Aware Network for Visual Tracking》(SANet) CVPR 2017 1FPS
（《Object tracking via dual linear structured SVM and explicit feature map》 CVPR 2016
ADNet《Action-Decision Net. for Tracking with Deep Reinforcement Learning》 CVPR 2017）
Siamese Network Based Tracker:
GOTURN 《Learning to track at 100 fps with deep regression networks ECCV, 2016》效果差 100FPS
(SINT《Siamese instance search for tracking》 CVPR 2016)
Graph Based Trackers:
TCNN 《 Modeling and propagating cnns in a tree structure for visual tracking》1.5fps 2016年VOT冠军
(《Superpixel-Based Tracking-By-Segmentation Using Markov Chains》 CVPR 2017 基于分割