Deep Meta Learning for Real-Time Visual Tracking based on Target-Specific Feature Space 

2018-01-04  15:58:15 

  

  写在前面:为什么要看这个paper?这篇 paper 貌似是第一个将 meta-learning 应用到 visual tracking 领域的,取得了速度和精度较好的平衡。

  Introduction:

  我们知道,tracking 中比较重要的就是 target object 特征的学习 以及 物体外观的变化。很多算法都针对这两点一直进行改进,而最近 NN 对特征的表达提供了很好的解决,但是,物体外观的变化,仍然不能很好的处理,很多都是 用跟踪的结果弄一个 target object 的集合,然后适时的进行更新。但是,这种策略是不可避免的,分类器通常都会 overfitting,然后丢失了 the generalization capabilities due to the insufficient training of samples.

  本文基于以上背景和动机,提出了一种 end to end visual tracking network structure,主要包括了两个部分:

  一个是:Siamese matching network for target search,

  另一个是:meta-learning network for adaptive feature space.

  这里我们主要关注的是这个 meta-learning network,我们提出的一个 参数预测网络(parameter prediction network),当然这里是借鉴了最新的 meta-learning 用于 few-shot learning problem.

  

  The proposed meta-learner network is trained to provide the matching network with additional convolutional kernels so that the feature space of the matching network can be modified adaptively to adopt new appearance templates obtained in the course of tracking. The meta-learner network only sees the gradients from the last layer of the matching network, given new training samples for the appearance.

  We also employ a novel training scheme for the meta-learner network to maintain the generalization capability of the feature space by preventing the meta-learner network from generating new parameters that causes overfitting of the matching network. By incorporating our metalearner network, the target-specific feature space can be constructed instantly with a single forward pass without any iterative computation and optimization and free-from the innate overfitting. Fig.1 illustrates the motivation of proposed visual tracking algorithm.

  

  

  Tracking with Meta-Learner

  1. Overview of Proposed Method

  1.1. Compoent

  本文所涉及到的网络结构有两个部分构成:the matching network and meta-learning network.

  Siamese Matching Network 用来计算两个 image patch 之间的相应图(the response map):

  

  这部分特征提取 CNN是 fully convolutional network,损失函数就是计算:预测的响应图 和  groundtruth Response map 的差异。

  

  Meta-learning Network:这个网络提供的是  the matching network with target-specific weights given an image patch of the target with context patches z = {z1, ..., zM}.

  为了调整 weights 超向 target patch,我们利用 损失函数的平均负梯度 $\delta$ 来更新 matching network 的最后一层:

  

  The meta-learning network 的设计是基于一个假设:the characteristic of $\delta$ is empirically different according to a target.  这句话是什么意思 ?

  

  然后,这里将 $\delta$ 作为输入,the meta-learning network $g_{\theta}(*)$ 对应输入的 target-specific weights $w^{weights}$:

  

  其中,$\theta$ 是 the meta-learning network 的参数。这个新的 weights 被用来更新 matching network's 原始权重:

  

  其中, 连接了 $w^{target}$ to $w_{N}$ of last layer for feature extraction. 本文方法的流程图,如图2所示。
  

  

  

  

  

  Experiment:

    

论文阅读之:Deep Meta Learning for Real-Time Visual Tracking based on Target-Specific Feature Space的更多相关文章

  1. 【论文阅读】Deep Mutual Learning

    文章:Deep Mutual Learning 出自CVPR2017(18年最佳学生论文) 文章链接:https://arxiv.org/abs/1706.00384 代码链接:https://git ...

  2. 读论文系列:Deep transfer learning person re-identification

    读论文系列:Deep transfer learning person re-identification arxiv 2016 by Mengyue Geng, Yaowei Wang, Tao X ...

  3. 论文笔记:Learning Attribute-Specific Representations for Visual Tracking

    Learning Attribute-Specific Representations for Visual Tracking AAAI-2019 Paper:http://faculty.ucmer ...

  4. 论文解读《Deep Resdual Learning for Image Recognition》

    总的来说这篇论文提出了ResNet架构,让训练非常深的神经网络(NN)成为了可能. 什么是残差? "残差在数理统计中是指实际观察值与估计值(拟合值)之间的差."如果回归模型正确的话 ...

  5. 论文笔记:Deep Residual Learning

    之前提到,深度神经网络在训练中容易遇到梯度消失/爆炸的问题,这个问题产生的根源详见之前的读书笔记.在 Batch Normalization 中,我们将输入数据由激活函数的收敛区调整到梯度较大的区域, ...

  6. 【文献阅读】Deep Residual Learning for Image Recognition--CVPR--2016

    最近准备用Resnet来解决问题,于是重读Resnet的paper <Deep Residual Learning for Image Recognition>, 这是何恺明在2016-C ...

  7. 【论文阅读】Deep Clustering for Unsupervised Learning of Visual Features

    文章:Deep Clustering for Unsupervised Learning of Visual Features 作者:Mathilde Caron, Piotr Bojanowski, ...

  8. 论文阅读:Deep Attentive Tracking via Reciprocative Learning

    Deep Attentive Tracking via Reciprocative Learning 2018-11-14 13:30:36 Paper: https://arxiv.org/abs/ ...

  9. 【论文阅读】Deep Mixture of Diverse Experts for Large-Scale Visual Recognition

    导读: 本文为论文<Deep Mixture of Diverse Experts for Large-Scale Visual Recognition>的阅读总结.目的是做大规模图像分类 ...

随机推荐

  1. Linux 查看磁盘或文件夹及文件大小

    当磁盘大小超过标准时会有报警提示,这时如果掌握df和du命令是非常明智的选择. df可以查看一级文件夹大小.使用比例.档案系统及其挂入点,但对文件却无能为力.    du可以查看文件及文件夹的大小. ...

  2. Sitecore详细安装(包含sitecore安装过程截图)

    一.到Sitecore 官网下载安装包 1)浏览器中输入https://dev.sitecore.net/Downloads/Sitecore_Experience_Platform.aspx 2)安 ...

  3. devstack 部署 openstack(pick/mitaka)

    链接出处: https://blog.csdn.net/qiqishuang/article/details/51990662 报错处理出处:https://blog.csdn.net/wang114 ...

  4. 20155228 获取技能的成功经验和关于C语言学习的调查

    内容提要 你有什么技能比大多人(超过90%以上)更好?针对这个技能的获取你有什么成功的经验?与老师博客中的学习经验有什么共通之处? 有关C语言学习的调查 你是怎么学习C语言的?(作业,实验,教材,其他 ...

  5. Spring 无缝整合 quartz

    关键步骤: 1. 配置 SchedulerFactoryBean <bean class="org.springframework.scheduling.quartz.Schedule ...

  6. js中利用cookie实现记住密码功能

    在登录界面添加记住密码功能,代码如下: //设置cookie var passKey = '4c05c54d952b11e691d76c0b843ea7f9'; function setCookie( ...

  7. Autel MaxiSys Elite Diagnostic Tool Common problem solving methods

    1. updating MaxiFlash Elite to firmware 3.21? My maxisys communicate with the MaxiFlash J2534 but Ma ...

  8. How to diagnose vehicle fault code by BMW ICOM and ISTA-D software

    Today illustrate how to diagnose vehicle fault code by BMW diagnostic tool BMW ICOM and ISTA-D softw ...

  9. 自学Java第三个星期的总结

    在这一周里我在网上学习了java的分支结构.Number&Matht类.Character类.string类.String Buffer和String Builder类以及数组和日期时间等有关 ...

  10. Zabbix客户端(被监控端)安装配置

    1) 创建用户 groupadd zabbix useradd -g zabbix zabbix 2)zabbix软件包下载,安装 zabbix-2.2.6 http://jaist.dl.sourc ...