CODE

所属分类:matlab编程
开发工具:matlab
文件大小:6756KB
下载次数:896
上传日期:2011-04-09 12:42:30
上 传 者zhangdada007
说明:  1.GeometricContext文件是完成图片中几何方向目标分类。 参考文献《Automatic Photo Pop-up》Hoiem 2005 2 GrabCut文件是完成图像中目标交互式分割 参考文献《“GrabCut” — Interactive Foreground Extraction using Iterated Graph Cuts》 C. Rother 2004 3 HOG文件是自己编写的根据HOG特征检测行人的matlab代码 4 虹膜识别程序是下载的一个通用的虹膜识别程序,可以运行 5 GML_AdaBoost_Matlab_Toolbox是一个很好用的adaboost matlab工具箱 6 libsvm-mat-2.91-1 是用C编写的改进的SVM程序,代码质量很高,提供了matlab接口 7 SIFT_Matlab 是编写的利用sift特征进行的宽基线匹配,代码质量高 8 FLDfisher 是利用fisher 线性降维方法进行人脸识别
(1.GeometricContext file is complete the picture in the geometric direction of target classification. References " Automatic Photo Pop-up" Hoiem 2005 2 GrabCut the target file is an interactive segmentation of image reference " " GrabCut " - Interactive Foreground Extraction using Iterated Graph Cuts" C. Rother 2004 3 HOG documents prepared under their own HOG Characteristics of pedestrian detection matlab code 4 iris recognition process is to download a general iris recognition program, you can run 5 GML_AdaBoost_Matlab_Toolbox is a good use of adaboost matlab toolbox 6 libsvm-mat-2.91-1 is written in C to improve the SVM procedures, code of high quality, provides a matlab interface to 7 SIFT_Matlab is prepared for the use of sift features a wide baseline matching, the code is the use of high quality 8 FLDfisher fisher linear dimension reduction method for face recognition)

文件列表:
上传代码\GeometricContext\LICENSE.txt (736, 2010-01-10)
上传代码\GeometricContext\test_dir\results\Thumbs.db (13312, 2011-04-09)
上传代码\GeometricContext\test_dir\results\tmpimsp8247727.g.png (147745, 2011-02-27)
上传代码\GeometricContext\test_dir\results\tmpimsp8247727.v.png (514676, 2011-02-27)
上传代码\GeometricContext\test_dir\images\alley01.jpg (110208, 2010-01-10)
上传代码\GeometricContext\test_dir\images\city10.jpg (61448, 2010-01-10)
上传代码\GeometricContext\test_dir\images\lakecomo2008.jpg (277512, 2008-07-10)
上传代码\GeometricContext\test_dir\images\M2U00037[(000047)17-22-11].JPG (74971, 2011-02-25)
上传代码\GeometricContext\test_dir\images\Thumbs.db (15360, 2011-02-27)
上传代码\GeometricContext\src\APPgetLabeledImageM2005.asv (1431, 2011-02-25)
上传代码\GeometricContext\src\APPgetLabeledImageM2005.m (1384, 2011-02-25)
上传代码\GeometricContext\src\APPtestDirectory.m (2962, 2010-01-10)
上传代码\GeometricContext\src\APPtestImage.m (5220, 2011-02-22)
上传代码\GeometricContext\src\APPtestImageM2005.asv (4822, 2011-02-25)
上传代码\GeometricContext\src\APPtestImageM2005.m (4784, 2011-02-25)
上传代码\GeometricContext\src\boost_classify.m (1100, 2011-02-25)
上传代码\GeometricContext\src\classifiers_08_22_2005.mat (231989, 2010-01-08)
上传代码\GeometricContext\src\Gclassify.mat (71436, 2011-02-25)
上传代码\GeometricContext\src\ijcvTestImage.m (4125, 2010-01-10)
上传代码\GeometricContext\src\ijcvTestImageList.m (1267, 2010-01-10)
上传代码\GeometricContext\src\im2superpixels.m (327, 2011-02-22)
上传代码\GeometricContext\src\mccExcludedFiles.log (374284, 2011-02-19)
上传代码\GeometricContext\src\msLabelMap2Sp.m (0, 2011-02-21)
上传代码\GeometricContext\src\photoPopup.m (3032, 2010-01-09)
上传代码\GeometricContext\src\photoPopupIjcv.m (3188, 2010-01-09)
上传代码\GeometricContext\src\photoPopupM.m (2967, 2011-02-21)
上传代码\GeometricContext\src\photoPopupM2005.asv (2216, 2011-02-27)
上传代码\GeometricContext\src\photoPopupM2005.m (2240, 2011-02-27)
上传代码\GeometricContext\src\runtrain.asv (5579, 2011-02-23)
上传代码\GeometricContext\src\runtrain.m (5704, 2011-02-24)
上传代码\GeometricContext\src\runtrainadaboost.asv (6673, 2011-02-25)
上传代码\GeometricContext\src\runtrainadaboost.m (6716, 2011-02-25)
上传代码\GeometricContext\src\segment.exe (490328, 2011-02-19)
上传代码\GeometricContext\src\segment_directory.pl (667, 2007-12-09)
上传代码\GeometricContext\src\vrml\APPcreateGroundPoints.m (588, 2010-01-09)
上传代码\GeometricContext\src\vrml\APPfitGroundHough.m (14453, 2010-01-09)
上传代码\GeometricContext\src\vrml\APPlabels2planes.m (12662, 2010-01-09)
上传代码\GeometricContext\src\vrml\APPplanes2faces.m (4779, 2010-01-09)
上传代码\GeometricContext\src\vrml\APPwriteVrmlModel.m (2780, 2011-02-27)
... ...

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% SOURCE README FOR AUTOMATIC PHOTO POPUP AND GEOMETRIC CONTEXT Derek Hoiem (dhoiem@cs.cmu.edu) 01/08/2010 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% LICENSE Copyright (C) 2007 Derek Hoiem, Carnegie Mellon University This software is available for only non-commercial use. See the attached license in LICENSE.txt. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% REFERENCES This code implements the following papers. Note that the implementation may not be exact. Please cite one or more of these papers, depending on the use. D. Hoiem, A.A. Efros, and M. Hebert, "Geometric Context from a Single Image", ICCV 2005. D. Hoiem, A.A. Efros, and M. Hebert, "Recovering Surface Layout from an Image", IJCV, Vol. 75, No. 1, October 2007. D. Hoiem, A.A. Efros, and M. Hebert, "Automatic Photo Pop-up", ACM SIGGRAPH 2005. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Note about VERSIONS This contains two versions of the "Automatic Photo Pop-up" and "Geometric "Context" code. The first version is from the SIGGRAPH 2005 / ICCV 2005 papers and involves the functions photoPopup and APPtestImage. The second version is from the IJCV 2007 paper and involves the functions photoPopupIjcv and ijcvTestImage. Note that these implementations may not be exact. In particular, some changes to features were made to improve speed, which may have small effects on accuracy. Further, the multiple segmentation algorithm is random, so different runs will not produce identical results. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% How to RUN: First, appropriately replace the path for the segment command in im2superpixels.m. Geometric Context version 1 (ICCV 2005): [labels, conf_map, maps, pmaps] = APPtestImage(image, imsegs, vClassifier, hClassifier, segDensity) image should be a double color image imsegs is the superpixel structure (if empty, will be computed) The remaining three arguments are stored in data/classifiers_08_22_2005.mat [labels, conf_map, imsegs] = ... APPtestDirectory(segDensity, vClassifier, hClassifier, imdir, imfn, varargin) Processes a directory of images with given filenames. Optional last argument is where to store displayed results. Geometric Context version 2 (IJCV 2007): [pg, data, imsegs] = ijcvTestImage(im, imsegs, classifiers, smaps, spdata, adjlist, edata); The only required fields are im and classifiers. Example usage: load '../data/ijcvClassifier.mat' [pg, data, imsegs] = ijcvTestImage(im, [], classifiers); Classifiers trained on indoor and outdoor images are provided in the data directory. [pg, smaps, imsegs] = ijcvTestImageList(fn, imsegs, classifiers, laboutdir, confoutdir) Processes all images whose names are given in fn. As above, imsegs can be empty. Automatic Photo Pop-up (SIGGRAPH 2005, IJCV 2007) photoPopup(fnData, fnImage, extSuperpixelImage, outdir) This is the 2005 version. fnData is the filename containing the classifiers (data/classifiers_08_22_2005.mat). If extSuperpixelImage is empty, it will be computed. photoPopupIjcv(fnData, fnImage, extSuperpixelImage, outdir) This is the IJCV 2007 version. See notes above. Training: For the IJCV version, see src/ijcv06/ijcvMultiSegScript.m This function performs both training and evaluation. It requires some edits. It is currently setup for cross-validation, but it should be straightforward to use separate train and test sets. The Geometric Context Dataset (separate download) contains rand_indices which specifies the train/test splits. Other useful functions: src/util/splitpg.m src/util/pg2confidenceImages.m src/util/pg2* src/util/write* %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% NOTES: Superpixels: I use the segmentation code provided by Felzenszwalb and Huttenlocher at people.cs.uchicago.edu/~pff/segment/ to create the superpixels in my experiments. The first three arguments (sigma, k, min) that I use are 0.8 100 100. I've included a pl script for segmenting a directory that you may find useful. You can also use a different program to create the superpixel image. That image should have a different RGB color for each segment without drawn boundaries between segments. (C) Derek Hoiem, Carnegie Mellon University, 2007

近期下载者

相关文件


收藏者