Part Number: TDA3XEVM
Tool/software: WEBENCH® Design Tools
Hi,Guys.I am using tda3x to do the Pedestrian Detection in TDA3x.I am trying with the adaboost with ACF.I have already sucessfully used the usecase you provide.And I replace the outcome with my own(produced by the matlab tool box from https://github.com/tidsp/acf-jacinto). My question is:
1
In acfJacintoTrainTest.m we have two parameters:opts.modelDs and opts.modelDsPad the default values are as follows:
%opts.modelDs=[100 41]; opts.modelDsPad=[128 64];
%opts.modelDs=[56 24]; opts.modelDsPad=[64 32];
opts.modelDs=[56 24]; opts.modelDsPad=[64 64];
If I change the opts.modelDsPad to some other value like [66 34],the dsp can not use the outcome of matlab.Is this common?If I want to change these parameters,how should I do?
2.
According to adaboost,every detected window(Pedestrian) has a score above 0(ignore the window with a score below 0),If I want to set the score above say 4 to be 0.By this means I want to reduce the false positive(I know I will also reduce the recall).I can do it in matlab in the detect process,but how should I do this process in TDA3x?Thank you!