Quantcast

Re: To train a better classifier

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: To train a better classifier

fractalities


http://tech.groups.yahoo.com/group/OpenCV/message/58615


--- In [hidden email], "zhangjin2272" <zhangjin2272@...> wrote:

>
> Hello All,
> Recently I am trying to train a classifier with it I can caculate
> how many people entered from a door.The first thing i have to do is
> finding lots of images as positive samples, and some images as  
> negative samples. As it is a fixed-scene, the negative sample
> (background) I just used one without object.
> Now the problem is I don't really know how I should set the
> parameters of the "Createsample.exe" routine and
> the "haartraining.exe" routine. Anyone with good recommendation will
> be useful,Thanks!!
>
> ##[-info <collection_file_name>]
> ##[-img <image_file_name>]
> ##[-vec <vec_file_name>]
> ##[-bg <background_file_name>]
> ##[-num <number_of_samples = 1000>]
> [-bgcolor <background_color = 0>]    //Should I set the bgcolor? How?
> [-inv] [-randinv] [-bgthresh <background_color_threshold = 80>]
> [-maxidev <max_intensity_deviation = 40>]  
> // Should I set it? how?
> [-maxxangle <max_x_rotation_angle = 1.100000>]
> // Should I set it? how?
>
> [-maxyangle <max_y_rotation_angle = 1.100000>]
> // Should I set it? how?
>
> [-maxzangle <max_z_rotation_angle = 0.500000>]
> // Should I set it? how?
>
> [-show [<scale = 4.000000>]]
> #[-w <sample_width = 24>]
> #[-h <sample_height = 24>]
>


Loading...