文件列表:

甚老师模型1\airTe2\image_0031.jpg (21175, 2004-11-09)
甚老师模型1\airTe2\image_0047.jpg (15692, 2004-11-09)
甚老师模型1\airTe2\image_0048.jpg (10456, 2004-11-09)
甚老师模型1\airTe2\image_0051.jpg (14739, 2004-11-09)
甚老师模型1\airTe2\image_0052.jpg (11925, 2004-11-09)
甚老师模型1\airTe2\image_0070.jpg (16455, 2004-11-09)
甚老师模型1\airTe2\image_0071.jpg (12193, 2004-11-09)
甚老师模型1\airTe2\image_0072.jpg (18748, 2004-11-09)
甚老师模型1\airTe2\image_0079.jpg (17754, 2004-11-09)
甚老师模型1\airTe2\image_0083.jpg (16543, 2004-11-09)
甚老师模型1\airTe3\image_0050.jpg (13811, 2004-11-09)
甚老师模型1\airTe3\image_0053.jpg (20985, 2004-11-09)
甚老师模型1\airTe3\image_0062.jpg (10075, 2004-11-09)
甚老师模型1\airTe3\image_0072.jpg (14511, 2004-11-09)
甚老师模型1\airTe3\image_0073.jpg (10739, 2004-11-09)
甚老师模型1\airTe3\image_0075.jpg (15549, 2004-11-09)
甚老师模型1\airTe3\image_0076.jpg (15043, 2004-11-09)
甚老师模型1\airTe3\image_0082.jpg (9832, 2004-11-09)
甚老师模型1\airTe3\image_0085.jpg (14245, 2004-11-09)
甚老师模型1\airTe3\image_0088.jpg (11209, 2004-11-09)
甚老师模型1\airTr1\image_0002.jpg (21202, 2004-11-09)
甚老师模型1\airTr1\image_0003.jpg (13909, 2004-11-09)
甚老师模型1\airTr1\image_0004.jpg (16181, 2004-11-09)
甚老师模型1\airTr1\image_0005.jpg (16297, 2004-11-09)
甚老师模型1\airTr1\image_0006.jpg (16463, 2004-11-09)
甚老师模型1\airTr1\image_0007.jpg (13989, 2004-11-09)
甚老师模型1\airTr1\image_0008.jpg (21144, 2004-11-09)
甚老师模型1\airTr1\image_0010.jpg (16400, 2004-11-09)
甚老师模型1\airTr1\image_0012.jpg (12130, 2004-11-09)
甚老师模型1\airTr1\image_0013.jpg (16091, 2004-11-09)
甚老师模型1\airTr2\image_0001.jpg (10798, 2004-11-09)
甚老师模型1\airTr2\image_0002.jpg (13444, 2004-11-09)
甚老师模型1\airTr2\image_0003.jpg (14218, 2004-11-09)
甚老师模型1\airTr2\image_0005.jpg (15632, 2004-11-09)
甚老师模型1\airTr2\image_0007.jpg (12178, 2004-11-09)
甚老师模型1\airTr2\image_0008.jpg (15824, 2004-11-09)
甚老师模型1\airTr2\image_0011.jpg (20042, 2004-11-09)
甚老师模型1\airTr2\image_0012.jpg (14987, 2004-11-09)
甚老师模型1\airTr2\image_0013.jpg (21837, 2004-11-09)
甚老师模型1\airTr2\image_0014.jpg (13816, 2004-11-09)
... ...

============================================================================== Copyright 2005 Center for Biological & Computational Learning at MIT and MIT All rights reserved. Permission to copy and modify this data, software, and its documentation only for internal research use in your organization is hereby granted, provided that this notice is retained thereon and on all copies. This data and software should not be distributed to anyone outside of your organization without explicit written authorization by the author(s) and MIT. It should not be used for commercial purposes without specific permission from the authors and MIT. MIT also requires written authorization by the author(s) to publish results obtained with the data or software and possibly citation of relevant CBCL reference papers. We make no representation as to the suitability and operability of this data or software for any purpose. It is provided "as is" without express or implied warranty. ============================================================================== This directory contains a new implementation of the methods described in: T. Serre, L. Wolf and T. Poggio. Object Recognition with Features Inspired by Visual Cortex. To Appear In: Proceedings of 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR 2005), IEEE Computer Society Press, San Diego, June 2005. A more complete AI Memo version is available at (ftp://publications.ai.mit.edu/ai-publications/2004/AIM-2004-026.pdf). This new implementation is faster and easier to use than the original one, and was tested to be almost fully compatible with it. Lior Wolf, Stan Bileschi and Thomas Serre contributed to this implementation which is based on the C implementation written by Thomas. Parts of that C implementation were based on an earlier C code by Max Riesenhuber, BEFORE YOU RUN ============== 1. Set the directories for the positive and negative images. There are four directories: train_set.pos, train_set.net, test_set.pos and test_set.net. These are set at lines 18-21 of demoRelease.m 2. The input images have to be grayscale. Although the algorithm has some invariance to scale, it might be a good idea to scale all images to the same height. We use a height of 140 pixels in many of our experiments. 3. If you would like to use an SVM classifier, please install OSU SVM (http://www.ece.osu.edu/~maj/osu_svm/). Add the path of OSU SVM to the second line of demoRelease. If you'd prefer a NN classifier set useSVM (line 5) to zero. 4. If you'd like the algorithm to learn its own object-specific features set READPATCHESFROMFILE (line 9) to zero (should give somewhat better results on many datasets). If not, you can use the universal features that are stored in PatchesFromNaturalImages250per4sizes.mat. This is done by setting READPATCHESFROMFILE to one. The stored prototypes are taken from a set of "natural images" we collected. Contents of the directory: MAIN DEMO ========= demoRelease.m - a demo showing a possible use of the code in this directory. The demo reads a set of images, extracts the C2 features for this set, and builds a classifier to learn the class of images. MAIN STANDARD MODEL FUNCTIONS ============================= C1.m - extracts s1 and c1 layers of the standard model representation C2.m - extracts s2 and c2 layers of the standard model representation CLASSIFICATION FUNCTIONS ======================== CLSnn.m - Nearest Neighbor classifier train CLSnnC.m - Nearest Neighbor classifier test CLSosusvm.m - SVM train (a wrapper function for osusvm) CLSosusvmC.m - SVM test (a wrapper function for osusvm) UTILITY FUNCTIONS ================= extractC2forcell.m - extracts C2 for all images in a cell, using all prototypes in another cell extractRandC1Patches.m - extracts random C1 patches to serve as prototypes init_gabor.m - creates Gabor filters maxfilter.m - local maximum in an image padimage.m - adds zeros around the boundary of the image readAllImages.m - reads all images in training and testing directories into one cell sumfilter.m - local sums in an image unpadimage.m - reverses the effect of padimage WindowedPatchDistance.m - scans an image looking for the best match for a prototype (image fragment) PRECOMPUTED UNIVERSAL FEATURES ============================== PatchesFromNaturalImages250per4sizes.mat %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% BUG FIXES %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Jul 01 2005 - changed bugs in extractC2forcell, extractRandC1Patches.m

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：