I agree that adaboost can only process uniform size of input
features.I wonder it ,too.
How about uniform these different features under the same framework?I
mean thant you can try to use some transform to make the two differnt
kind of features use the same comparing rule .For example,if you use
lbp transform to the blocks,the features can be uniformed into 59 or
58 bin,then we can implement same rules to form weak classifiers.
On 5月26日, 上午8时34分, ligang zhang <zlg...@gmail.com> wrote:
> Hi Everyone,
> Currently, I try to use Adaboost for salient features selection. I encounter
> one problem: if the input features have different sizes (e.g 4*4, 8*8
> block), how to implement the Adaboost? I read the code of Adaboost, and find
> that it seems only suitable for uniform size input features.
> Thanks very much.
> Li
|
