文件列表:

.saic.ini.swp (12288, 2004-10-07)
a.out (43147, 2004-10-07)
affine.h (2908, 2004-10-06)
image.h (1706, 2004-10-04)
random.h (652, 2004-09-28)
readINI.h (1493, 2004-10-06)
saic.ini (230, 2004-10-06)
saicdataset1 (1403, 2004-10-06)
svdcmp.h (3963, 2004-09-28)
tracker.c++ (2005, 2004-10-07)
tracker.h (14524, 2004-10-07)
trackingIPTrans.pdf (1254899, 2004-10-07)
vector.h (1126, 2004-09-28)
result\output.txt (4757, 2004-10-07)
result (0, 2004-10-07)
saic\mi1A_YDB0170.pgm (86415, 2004-10-04)
saic\mi1A_YDB0171.pgm (86415, 2004-10-04)
saic\mi1A_YDB0172.pgm (86415, 2004-10-04)
saic\mi1A_YDB0173.pgm (86415, 2004-10-04)
saic\mi1A_YDB0174.pgm (86415, 2004-10-04)
saic\mi1A_YDB0175.pgm (86415, 2004-10-04)
saic\mi1A_YDB0176.pgm (86415, 2004-10-04)
saic\mi1A_YDB0177.pgm (86415, 2004-10-04)
saic\mi1A_YDB0178.pgm (86415, 2004-10-04)
saic\mi1A_YDB0179.pgm (86415, 2004-10-04)
saic\mi1A_YDB0180.pgm (86415, 2004-10-04)
saic\mi1A_YDB0181.pgm (86415, 2004-10-04)
saic\mi1A_YDB0182.pgm (86415, 2004-10-04)
saic\mi1A_YDB0183.pgm (86415, 2004-10-04)
saic\mi1A_YDB0184.pgm (86415, 2004-10-04)
saic\mi1A_YDB0185.pgm (86415, 2004-10-04)
saic\mi1A_YDB0186.pgm (86415, 2004-10-04)
saic\mi1A_YDB0187.pgm (86415, 2004-10-04)
saic\mi1A_YDB0188.pgm (86415, 2004-10-04)
saic\mi1A_YDB0189.pgm (86415, 2004-10-04)
saic\mi1A_YDB0190.pgm (86415, 2004-10-04)
saic\mi1A_YDB0191.pgm (86415, 2004-10-04)
saic\mi1A_YDB0192.pgm (86415, 2004-10-04)
saic\mi1A_YDB0193.pgm (86415, 2004-10-04)
... ...

Video Tracker Version 0.0.1 --------------------------- Do not distribute without prior approval from Prof. Rama Chellappa, University of Maryland, College Park (email: rama@cfar.umd.edu) This code is based on the "Visual tracking and recognition using appearance-adaptive models in particle filters" (IEEE transactions on Image Processing, Nov 2004) by Shaohua Zhou, Rama Chellappa and Baback Moghaddam. A copy of this paper can be found with this tar file. You can contact Aswin C Sankaranarayanan (aswch@cfar.umd.edu) for any doubts/clarifications regarding this code. C-Code ------ Files: tracker.h -- Contains all tracker related functions vector.h -- defines basic vector unit with element operations affine.h -- defines affine structure with affine operations image.h -- image operations random.h -- gaussian noise generator svdcmp.h -- Modified NR code for SVD readINI.h -- reads INI file README -- this document tracker.c++ -- wrapper/interface code to read ini-files/image data and run tracker result/ ---> folder in which results are dumped COMPILING --------- All header files are linked thru the C++ file. A simple compile should do the trick c++ tracker.c++ CONFIGURATION FILES ------------------- to run a tracker over a dataset, 2 files need to be created 1) ini file: see example: saic.ini file 2) image-list file: lists all the images over which tracker needs to be run (see saicdataset1 file for format) 2 examples with INI files etc can be found (see saic.ini and car.ini) The ini file structure is described below file_path= number_of_samples = template_size = ---- it is reccomended to keep aspect ratio of this template similar to that of target + number of pixels in the template > 2*number of samples (i.e the product of rows *col (of template ) > 2*number of samples) rel_std_afnv = [ affine noise std. variances ... leave at value used in examples ini files ] initial box values = [ initial box values in c1, r1, c2, r2, c3, r3 format (see below) ] number_of_frames = [ number of frames] movie = 0 --> no image output, 1-> image output in result/...../ folder save to disk = 0 --> dont save result. 1 --> save output in result/output.txt Further, tracker.c++ (line number 18) should be modified to point ot the ini file. OUTPUT ------ Text output is dumped in the results/output.txt file. Each line has the file name and the affine parameters of the MAP particle. Image output is saved in results/(filename as in image-list file) THINGS TO REMEMBER ------------------ 1)) The code as it exists now (0.0.1) only takes in GRAYSCALE PGM (raw) format w/o any remarks. Saving an image from matlab as an PGM file should work. IMAGEmagick toolkit (in linux) provides a commandline conversion utility called "convert" --- works most of the time. The basic format is P5 255 {ROW SCANNED pixel intensities as char (-128 to 128) binary format} 2)) The options "save to disk" controls if text output results/output.txt is created and "movie" determines if "burned" images are dumped. 3)) if "movie" option is selected, u may need to create appropriate folder in the result directory. (for ex, because all image names in saicdataset1 are saic/ ... we need a "saic" folder in result folder). 4)) IMPORTANT: WORKS ONLY IN LINUX (because of the way the code is written). U MAY WANT TO CONTACT Mr. Yang Ran (rany@cfar.umd.edu) for a version that works in Microsoft VC++.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：