Re3 : Real-Time Recurrent Regression Networks for Visual Tracking of Generic Objects

2019-10-04 14:42:54

Paper：https://arxiv.org/pdf/1705.06368.pdf

Code（TensorFlow）：https://gitlab.com/danielgordon10/re3-tensorflow

Demo Video: https://www.youtube.com/watch?v=RByCiOLlxug&feature=youtu.be

Related Trackers: GOTURN [Blog]

This paper is developed based on deep regression network, the key idea is to utilize the LSTM network to memorize the history information (i.e. the tracking results). In addition, they also utilize multi-layer's features for better target representation. The overall pipeline can be found in following figure:

 

This tracker work well, but the training time need to take about one week (7 days), as it mentioned in this paper. It is because the original deep regression network is already hard to train, and the introduced LSTM further increased the difficulty. The efficiency is really fast, 150 FPS on a GPU.

The imporvement based on GOTURN is significant, according to their experiments, as shown in Fig. 4.

Although this tracker seems simple, but it really works well in some videos and fast enough for practical applications, such as uav, robotic.

Re3 : Real-Time Recurrent Regression Networks for Visual Tracking of Generic Objects的更多相关文章

  1. 论文笔记之：Learning Multi-Domain Convolutional Neural Networks for Visual Tracking

    Learning Multi-Domain Convolutional Neural Networks for Visual Tracking CVPR 2016 本文提出了一种新的CNN 框架来处理 ...

  2. 论文笔记之：Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning

    论文笔记之:Action-Decision Networks for Visual Tracking with Deep Reinforcement Learning  2017-06-06  21: ...

  3. Paper Reading - Long-term Recurrent Convolutional Networks for Visual Recognition and Description ( CVPR 2015 )

    Link of the Paper: https://arxiv.org/abs/1411.4389 Main Points: A novel Recurrent Convolutional Arch ...

  4. 论文阅读：MDNet: Learning Multi-Domain Convolutional Neural Networks for Visual Tracking

    前言 CVPR2016 来自Korea的POSTECH这个团队   大部分算法(例如HCF, DeepLMCF)只是用在大量数据上训练好的(pretrain)的一些网络如VGG作为特征提取器,这些做法 ...

  5. Summary on Visual Tracking: Paper List, Benchmarks and Top Groups

    Summary on Visual Tracking: Paper List, Benchmarks and Top Groups 2018-07-26 10:32:15 This blog is c ...

  6. （转）CVPR 2016 Visual Tracking Paper Review

    CVPR 2016 Visual Tracking Paper Review  本文摘自:http://blog.csdn.net/ben_ben_niao/article/details/52072 ...

  7. 论文笔记之：Spatially Supervised Recurrent Convolutional Neural Networks for Visual Object Tracking

    Spatially Supervised Recurrent Convolutional Neural Networks for Visual Object Tracking  arXiv Paper ...

  8. 论文笔记：Learning regression and verification networks for long-term visual tracking

    Learning regression and verification networks for long-term visual tracking 2019-02-18 22:12:25 Pape ...

  9. Attention and Augmented Recurrent Neural Networks

    Attention and Augmented Recurrent Neural Networks CHRIS OLAHGoogle Brain SHAN CARTERGoogle Brain Sep ...

随机推荐

  1. 25、vuex改变store中数据

    以登录为例: 1.安装vuex:npm install vuex --save 2.在main.js文件中引入: import store from '@/store/index.js'new Vue ...

  2. 地产propretie单词propretie财产

    中文名:房产财产地产 外文名:property.propretie 释义:财产.所有物等 用法:作名词. 词汇搭配动词+-等 目录 1 英文释义 2 释义例句 3 词汇搭配 4 衍生 英文释义 1. ...

  3. Android 使用traceView

    Android在做性能优化的时候需要使用traceView进行检测,traceView可以详细的记录下线程执行的时间让我们在做优化的时候可以清楚优化哪些内容.首先我们需要使用这个traceView,在 ...

  4. elasticsearch使用ansj分词器

    目前elasticsearch的版本已经更新到7.0以上了,不过由于客户需要5.2.2版本的elasticsearch,所以还是需要安装的,并且安装上ansj分词器.在部署ES的时候,采用容器的方式进 ...

  5. (十二)Kubernetes 认证、授权与准入控制

    访问控制概述 API Server作为Kubernetes集群系统的网关,是访问和管理资源对象的唯一入口:包括kube-controller-manager.kube-scheduler.kubele ...

  6. Bag of Tricks for Image Classification with Convolutional Neural Networks笔记

    以下内容摘自<Bag of Tricks for Image Classification with Convolutional Neural Networks>. 1 高效训练 1.1 ...

  7. Linux四剑客之awk命令

    AWK详解   awk简介 awk其名称得自于它的创始人 Alfred Aho .Peter Weinberger 和 Brian Kernighan 姓氏的首个字母.实际上 AWK 的确拥有自己的语 ...

  8. Ruby for

    #!/usr/bin/ruby -w# -*- coding: UTF-8 -*-for i in 1..5 print i," "endprint "\n"f ...

  9. Content-Type与MIME

    http://www.cnblogs.com/jsean/articles/1610265.html 首先,我们要了解浏览器是如何处理内容的.在浏览器中显示的内容有 HTML.有 XML.有 GIF. ...

  10. SpringBoot 之Spring Boot Starter依赖包及作用（自己还没有看）

    spring-boot-starter 这是Spring Boot的核心启动器,包含了自动配置.日志和YAML. spring-boot-starter-amqp 通过spring-rabbit来支持 ...