文件列表:

Detector_demo (0, 2009-09-16)
Detector_demo\bb_burn.m (501, 2008-07-21)
Detector_demo\detectors (0, 2009-09-16)
Detector_demo\detectors\frontal.mat (3904751, 2008-07-25)
Detector_demo\detectors\info.txt (101, 2008-08-31)
Detector_demo\detectors\profile.mat (767925, 2009-06-14)
Detector_demo\fliplr_model.m (1453, 2008-03-11)
Detector_demo\go_offline.m (828, 2009-09-16)
Detector_demo\go_online.m (662, 2008-08-31)
Detector_demo\images (0, 2009-09-16)
Detector_demo\images\afghanistan.pgm (89185, 2008-01-12)
Detector_demo\images\life2100.pgm (76815, 2007-08-08)
Detector_demo\mex (0, 2009-09-16)
Detector_demo\mex\abFindObject.dll (94208, 2008-07-21)
Detector_demo\mex\abFindObject.mexa64 (846238, 2008-07-23)
Detector_demo\mex\abFindObject.mexglx (777286, 2008-07-23)
Detector_demo\paper (0, 2009-09-16)
Detector_demo\paper\bmvc2008.pdf (251188, 2008-07-23)
Detector_demo\usbclear.m (26, 2008-07-21)
Detector_demo\usbinit.m (66, 2008-08-31)

Running: --------------------------------------------------------------------------------------- > usbinit initialized Matlab Image Acquisition toolbox, windows alternative: VCAPG2, http://www.ikko.k.hosei.ac.jp/~matlab/matkatuyo/vcapg2.htm > go_online runs detector > usbinit clears the video stream Testing files: --------------------------------------------------------------------------------------- go_offline ... loads two images and tests frontal and profile detector go_online ... runs real-time detector on a input from a webcam requires: Image Acquisition Toolbox Detector description: --------------------------------------------------------------------------------------- Detector is implemented as a mex function (compiled for Windows, Linux 32/***bit) and called by the following function: [bb, conf] = abFindObject(img,model,scan); INPUT: img ... gray scale image model . detector scan .. scan.win_height = 30; % minimal bb height scan.win_width = 30; % minimal bb width scan.row_shift = 3; % row shift of window, not pixels! scan.col_shift = 3; % col shift of window, not pixels! scan.scale = 1.40; % scale change scan.threshold = 7; % minimal threshold returned by detector, scan.min_std = 1; % minimal std of patch that is considered as face, scan.nms = 1; % non maxima suppression, kills multiple detections scan.overlap = .4; % radius of bb considered as the same spot for NMS OUTPUT: bb .... bounding boxes of found objects conf .. confidences (log likelihood ratio) Please cite the following paper if you decide to use the code: ---------------------------------------------------------------------------------------- Zdenek Kalal, Jiri Matas, Krystian Mikolajczyk. Weighted Sampling for Large-Scale Boosting, BMVC, 2008

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：