DSST: Discriminative Scale Space Tracker
-------------------------------------------------------------------------------
DSST is part of the [Visual Object Tracking Repository](https://github.com/gnebehay/VOTR),
which aims at providing a central repository for state-of-the-art tracking algorithms that are freely available.
The source code for this tracker was submitted for the [VOT2014 challenge](https://www.votchallenge.net).
A minor adaption was made by using MATLAB's imresize() function
instead of a custom mex file to increase portability.
Dependencies:
* Matlab
* Image Processing Toolbox
* Signal Processing Toolbox
The following description was copied literally from the original author.
ABSTRACT
----------------------------------------------------------------------------
The Discriminative Scale Space Tracker (DSST), proposed in [2], extends the Minimum Output Sum of Squared
Errors (MOSSE) tracker [1] with robust scale estimation. The MOSSE tracker works by training a discriminative
correlation filter on a set of observed sample grayscale patches. This correlation filter is then applied to estimate the
target translation in the next frame. The DSST additionally learns a one-dimensional discriminative scale filter, that
is used to estimate the target size. The scale filter is trained by extracting several sample patches at different scales
around the current target position in the image. Each sample is represented by a fixed-length feature vector based on
HOG. These samples are used to learn a multi-channel one-dimensional discriminative filter for scale estimation. This
scale filter is generic and can be combined with any tracker that is limited to only estimating the target translation.
Given a new image, the DSST first applies a translation filter to obtain the most probable target location. The scale
filter is then applied at this location to estimate the target size. The tracking model is then updated with the new
information information of the target and background appearance. For the translation filter, we combine the intensity
features employed in the MOSSE tracker with a pixel-dense representation of HOG-features.
[1] D. S. Bolme, J. R. Beveridge, B. A. Draper, and Y. M. Lui. Visual object tracking using adaptive correlation filters. In CVPR,
2010.
[2] M. Danelljan, G. Häger, F. Shahbaz Khan, and M. Felsberg. Accurate scale estimation for robust visual tracking. In Proceedings
of the British Machine Vision Conference (BMVC), 2014.
README
----------------------------------------------------------------------------
This MATLAB code integrates the Discriminative Scale Space Tracker (DSST) [1] into the VOT 2014 evaluation kit. The implementation is built upon the code provided by [2]. The code provided by [3] is used for computing the HOG features.
Instructions:
Use the wrapper function "wrapper.m" without trax.
The code may be publicly available from the VOT homepage.
Contact:
Martin Danelljan
martin.danelljan@liu.se
[1] Martin Danelljan, Gustav Häger, Fahad Shahbaz Khan and Michael Felsberg.
"Accurate Scale Estimation for Robust Visual Tracking".
Proceedings of the British Machine Vision Conference (BMVC), 2014.
[2] J. Henriques, R. Caseiro, P. Martins, and J. Batista.
"Exploiting the circulant structure of tracking-by-detection with kernels."
In ECCV, 2012.
[3] Piotr Dollár.
"Piotr's Image and Video Matlab Toolbox (PMT)."
http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html.
没有合适的资源?快使用搜索试试~ 我知道了~
论文《Accurate scale estimation for robust visual tracking》代码
共13个文件
m:7个
hpp:2个
mexw64:1个
4星 · 超过85%的资源 需积分: 15 56 下载量 141 浏览量
2015-06-09
17:15:25
上传
评论 2
收藏 42KB ZIP 举报
温馨提示
论文《Accurate scale estimation for robust visual tracking》简单代码,来自VOT提供的代码,已经稍作修改,可以单独运行。 注意:修改vot_tracker_initialize.m文件中的base_path 才能运行,运行数据采用CVPR13上的benchmark。
资源推荐
资源详情
资源评论
收起资源包目录
DSST.zip (13个子文件)
DSST
get_feature_map.m 456B
vot_tracker_results.m 629B
wrappers.hpp 2KB
gradientMex.mexw64 30KB
wrapper.m 7KB
get_subwindow.m 2KB
gradientMex.cpp 18KB
vot_tracker_initialize.m 3KB
gradientMex.mexa64 23KB
get_scale_subwindow.m 1KB
fhog.m 3KB
sse.hpp 3KB
README.md 3KB
共 13 条
- 1
资源评论
- v7512152015-11-13非常好的代码
- u0104636342016-01-12代码可以运行,但是修改后运行速度变慢
- topsir2016-11-12非常好的学习资料
- Hank@Linux:~$love2019-03-26代码配置环境不同
TJXUNwu
- 粉丝: 45
- 资源: 15
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功