I have not yet found reading method for xml file created by opencv_traincascade LBP.
I am making cascade program to detect vehicle using opencv_traincascade attached in OpenCV2.0. I am suffering from xml file handling.
[Developing Environment] OpenCV2.0 Visual C++ 2008 Windows 7
Face detection program with OpenCV -------------------------(1)
http://www.aianet.ne.jp/~asada/prog_d...
I am trying to detect vehicle using static object detection program (1) and the following (2)-(4) xml file.
xml file created by opencv_haartraining --------------------(2)
xml file created by opencv_traincascade HARR ---------------(3)
xml file created by opencv_traincascade LBP ----------------(4)
I examined xml file by text editor and found that (2)-(4) format were different.
When I checked vehicle detection program (1) using (2)-(4) xml file,
Vehicle detection program run correctly with xml file (2) created by opencv_haartraining. Vehicle detection program did not run correctly with xml file (3) created by opencv_traincascade HARR. Vehicle detection program did not run correctly with xml file (4) created by opencv_traincascade LBP.
an error message for (3),(4) appeared as follows.
Non handling Exception by 0x75dc4503 was occurred. Microsoft C++ Exception : cv::Exception
If -baseFormatSave was added to (3) excecution option, new format xml file would be converted to old format xml file (compatible with (2)). But it took me 6 hours to compensate operation. I had to add stage No, tree No, and root node.
I have been searching solution method for about one month by Internet. But I have not yet found reading method for xml file (4) created by opencv_traincascade LBP.
I need help.
the code in your link is unfortunately horribly outdated, you must not use it. also, opencv2.0 is like 6 years old, please update to a more recent version.
the old CvHaarClassifierCascade (c-api) thing cannot read files created by opencv_traincascade, only files created from the deprecated opencv_haartraining tool.
http://docs.opencv.org/ref/master/d2/...
Also, in the last months we fixed several bugs in the existing traincascade interface, so please make sure you use at least the latest 2.4 branch from github!