hog

所属分类:图形图像处理
开发工具:matlab
文件大小:103KB
下载次数:1093
上传日期:2008-06-22 21:24:55
上 传 者matterhorn2007
说明:  hog特征提取算法的实现,用于object detection,特别是human detection,针对64*128的图像。
(hog feature extraction algorithm for object detection, in particular, human detection, against 64* 128 images.)

文件列表:
phog\anna_binMatrix.m (1183, 2008-03-04)
phog\anna_phog.m (1538, 2008-03-04)
phog\anna_phogDescriptor.m (1097, 2008-03-04)
phog\anna_phog_demo.m (109, 2008-05-29)
phog\anna_phog_demo.m.bak (109, 2008-03-04)
phog\image_0058.jpg (14200, 2004-11-09)
phog\phog\a.txt (12, 2008-06-03)
phog\phog\anna_binMatrix.asv (1196, 2008-05-29)
phog\phog\anna_binMatrix.m (1183, 2008-03-04)
phog\phog\anna_phog.asv (1298, 2008-05-29)
phog\phog\anna_phog.m (1288, 2008-05-29)
phog\phog\anna_phogDescriptor.m (499, 2008-05-29)
phog\phog\anna_phog_demo.asv (617, 2008-05-29)
phog\phog\anna_phog_demo.m (621, 2008-05-29)
phog\phog\crop.bmp (24630, 2008-05-28)
phog\phog\crop.bmp.txt (2, 2008-06-03)
phog\phog\example.asv (169, 2008-06-03)
phog\phog\example.m (176, 2008-06-03)
phog\phog\hog.asv (1135, 2008-06-02)
phog\phog\hog.m (1135, 2008-06-02)
phog\phog\hs_err_pid208.log (16369, 2008-05-26)
phog\phog\image_0058.jpg (14200, 2004-11-09)
phog\phog\image_0058.jpg.txt (166150, 2008-06-02)
phog\phog (0, 2008-06-06)
phog (0, 2008-06-03)

Using the Caltech 101 splits provided on the web and an SVM Classifier you should have the following results: Shape 180 Shape180 Shape360 Shape360 (whole image) (roi) (whole image) (roi) l=0 13.20 14.70 16.20 17.57 l=1 33.07 38.56 36.14 44.37 l=2 46.99 58.76 47.25 61.69 l=3 48.43 61.04 49.28 61.63 merge (exhaustive search) 51.76 ***.31 54.33 66.27 merge (varma method) 49.21 60.19 50.00 61.04 NOTE: SVM 1-vs-all has been used to learn the level weights when merging all the pyramid levels. ------------------------------------------------ Using the Caltech 256 splits provided on the web and an SVM Classifier you should have the following results: Shape 180 Shape180 Shape360 Shape360 (whole image) (roi) (whole image) (roi) merge (exhaustive search) 16.81 24.28 19.31 27.17 NOTE: again, SVM 1-vs-all has been used to learn the level weights when merging all the pyramid levels.

近期下载者

相关文件


收藏者