Re: BUG? OpenMP, GCC and Haar Cascade recognition.

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

Re: BUG? OpenMP, GCC and Haar Cascade recognition.

wl2776
Administrator
 
> BUT (yes, is a big BUT): It is showing MANY false-positives... much
> more than when using bad parameters in single-threaded mode... I can
> make the detection a little more robust by setting more strict
> parameters, but still there are quite a few false-positives popping in
> and out.
I can confirm this.
I have seen false positives too, when openmp was switched on.
 
> So my question: Is this a bug? Or is GCC not supported at all with


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: BUG? OpenMP, GCC and Haar Cascade recognition.

garethedwards34
--- In [hidden email], "wl27761" <wl2776@...> wrote:
>
>  
> > BUT (yes, is a big BUT): It is showing MANY false-positives...
much
> > more than when using bad parameters in single-threaded mode... I
can
> > make the detection a little more robust by setting more strict
> > parameters, but still there are quite a few false-positives
popping in
> > and out.
> I can confirm this.
> I have seen false positives too, when openmp was switched on.
>  
> > So my question: Is this a bug? Or is GCC not supported at all with
>

Version 1.1.0 is producing many more false-positives than V1.0.0 and
is MUCH slower (presumably not pruning enough FPs at early stages of
cascade. I've done extensive tests and rebuilds with various compilers
and it's definitely a bug....don't know *what* the bug is yet though
:(

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: BUG? OpenMP, GCC and Haar Cascade recognition.

garethedwards34
In reply to this post by wl2776
--- In [hidden email], "wl27761" <wl2776@...> wrote:
>
>  
> > BUT (yes, is a big BUT): It is showing MANY false-positives...
much
> > more than when using bad parameters in single-threaded mode... I
can
> > make the detection a little more robust by setting more strict
> > parameters, but still there are quite a few false-positives
popping in
> > and out.
> I can confirm this.
> I have seen false positives too, when openmp was switched on.
>  
> > So my question: Is this a bug? Or is GCC not supported at all with
>

Definite bug - not just GCC either. Turning on OpenMP generates lots
more false positives and makes it much slower



Loading...