Ask Your Question

flammxy1's profile - activity

2013-03-03 16:22:55 -0500 received badge  Student (source)
2013-01-07 13:49:09 -0500 asked a question Is LBP better than HOG in Human detection?

like bicyclist? pedestrians?

Because I read some papers about semantic lbp, fourier lbp is better than HOG and cov? How about integral HOG?

2012-10-27 17:55:39 -0500 asked a question Question about the describe parameter of the files used in performance when doing haartraining?

I am doing the performance.exe and I have done the haartraining.exe. So what about the describe of the files used in performance? Is that shoud like the positive describe file like "a.jpg 1 0 0 30 30" ? Or just be "a.jpg"?

If should be the former, in this describe the "0 (x), 0 (y)"and the "30 (w), 30 (h)" should be the size of postive samples used in createsample.exe? or should be the object detection file, like a target picture, "a.jpg 1 100 150 300 400" ?

If should be the latter, Why I can not detect anything shoud as a red rectangle? Actually when the latter, the performance does nothing.

And I have tried to modify the number in "a.jpg 1 0 0 30 30" to any kind of "a.jpg 1 x x xxx xxx" It can detect the object but every time the same result.

Does anyone have some experience about this?