BINGObjectness-master
所属分类:Windows编程
开发工具:matlab
文件大小:831KB
下载次数:16
上传日期:2017-11-24 15:37:40
上 传 者:
sherry3255
说明: matlab实现程明明的BING算法,对于后续比如目标检测、目标识别具有很大参考价值。
(Matlab implementation of Cheng Mingming's BING algorithm.It has great reference value for the following target detection and target recognition.)
文件列表:
000001.jpg (78771, 2016-08-19)
BINGMultiple.cpp (2331, 2016-08-19)
BINGMultiple.mexa64 (348418, 2016-08-19)
BINGSingle.cpp (2129, 2016-08-19)
BINGSingle.mexa64 (348254, 2016-08-19)
Example_BINGMultiple.m (155, 2016-08-19)
Example_BINGSingle.m (974, 2016-08-19)
Example_trainBING.m (126, 2016-08-19)
MxArray.cpp (19293, 2016-08-19)
MxArray.hpp (36883, 2016-08-19)
Objectness (0, 2016-08-19)
Objectness\LICENSE (1505, 2016-08-19)
Objectness\Src (0, 2016-08-19)
Objectness\Src\CMakeLists.txt (1344, 2016-08-19)
Objectness\Src\CMakeLists.txt.user (11646, 2016-08-19)
Objectness\Src\CmFile.cpp (4806, 2016-08-19)
Objectness\Src\CmFile.h (2718, 2016-08-19)
Objectness\Src\CmShow.cpp (2422, 2016-08-19)
Objectness\Src\CmShow.h (258, 2016-08-19)
Objectness\Src\CmTimer.h (2319, 2016-08-19)
Objectness\Src\DataSetVOC.cpp (6674, 2016-08-19)
Objectness\Src\DataSetVOC.h (2370, 2016-08-19)
Objectness\Src\FilterTIG.cpp (2794, 2016-08-19)
Objectness\Src\FilterTIG.h (1752, 2016-08-19)
Objectness\Src\ImgContrastBB.h (1723, 2016-08-19)
Objectness\Src\LibLinear (0, 2016-08-19)
Objectness\Src\LibLinear\LibLinear.vcxproj (4226, 2016-08-19)
Objectness\Src\LibLinear\blas (0, 2016-08-19)
Objectness\Src\LibLinear\blas\Makefile (293, 2016-08-19)
Objectness\Src\LibLinear\blas\blas.h (702, 2016-08-19)
Objectness\Src\LibLinear\blas\blasp.h (16460, 2016-08-19)
Objectness\Src\LibLinear\blas\daxpy.c (1205, 2016-08-19)
Objectness\Src\LibLinear\blas\ddot.c (1211, 2016-08-19)
Objectness\Src\LibLinear\blas\dnrm2.c (1306, 2016-08-19)
Objectness\Src\LibLinear\blas\dscal.c (1035, 2016-08-19)
Objectness\Src\LibLinear\linear.cpp (55369, 2016-08-19)
Objectness\Src\LibLinear\linear.h (2008, 2016-08-19)
... ...
# BING Objectness
BING Objectness proposal estimator Matlab (mex-c) wrapper, runs at 250
FPS at a i7 CPU (2.93Hz) with Ubuntu 12.04 ***-bit and Matlab R2013a.
## Introduction
This is the matlab wrapper of BING Objectness for efficient
objectness proposal estimator following the CVPR 2014 paper BING, please
consider to cite and refer to this paper.
@inproceedings{BingObj2014,
title={{BING}: Binarized Normed Gradients for Objectness Estimation at
300fps},
author={Ming-Ming Cheng and Ziming Zhang and Wen-Yan Lin and Philip H.
S. Torr},
booktitle={IEEE CVPR},
year={2014},
}
The original author Ming-Ming Cheng has already released the source code
for windows ***-bit platform, and Shuai Zheng has provided the code for
the linux/mac/windows users.
In this library, I intend to provide some simple functions so that
users in Matlab can easily reproduce the results in the paper or use BING for some
other applications.
## Requirements
In order to make the code running, you need to download the
images/annotations PASCAL VOC 2007 data from the link:
[VOC2007](http://pascallin.ecs.soton.ac.uk/challenges/VOC/voc2007/#testdata).
Please refer to the FAQs #2 in
[http://mmcheng.net/bing/](http://mmcheng.net/bing/) for more details
about how to prepare for the VOC2007 dataset.
## HowTo in Matlab
I have written three mex-c fuctions:
* trainBING.cpp - For training BING Objectness
* BINGMultiple.cpp - For reproducing the results of the original paper
* BINGSingle.cpp - For users to use BING on a single image
Also, I have written three matlab scripts to show how to use them in
Matlab (Name Convention: Example\_ + function name ).
Note that the function __trainBING__ is a bit needless because
__BINGMultiple__ and __BINGSingle__ themselves will learn the models if they do not
exists.
You can run Example\_BINGMultiple.m to reproduce the results in the
origin paper, and a script called __PerImgAll.m__ will be generated in
your VOC2007 folder.
You can use the script as well as __PlotsCVPR14.m__ to plot the Figure 3 in the
paper.
I have tested the code in Ubuntu 12.04 ***-bit (8G Memory) and Matlab R2013a,
and it produces the same accuarcy results as the
original windows version, except that in my PC, it runs at 250 FPS
compared to 300 FPS reported in the paper.
If you intend to use the codes in other system version, please run
_compile.m_ to re-compile the files.
Please contact me (removethisifyouarehuman-tfzhou@bit.edu.cn) or create an issue if you have problems to run the
codes.
## Other Source Code Repos
* [http://mmcheng.net/bing/](http://mmcheng.net/bing/) -- The original
code / FAQ / Paper by Ming-Ming Cheng.
* [https://github.com/bittnt/Objectness](https://github.com/bittnt/Objectness) -- The Linux version library by Shuai Zheng.
## License
BSD license.
近期下载者:
相关文件:
收藏者: