Face Detect Engine with LB365 - Leopardboard 365/368 - Leopardboard

Hi,
i work with the LB 365 and i try to get the Face Detect Engine running with the latest RR SDK. Below is the link where you can find my code for the face detect engine, which i took out of the FD engine data sheet (http://www.ti.com/lit/ug/sprugg8a/sprugg8a.pdf).

http://pastebin.com/EE4GuyL5

Location:~/RRSDK/kernel/linux-2.6.32.17-psp03.01.01.39/drivers/char File:dm365_ipipe_hw.c Line:483
static void rsz_set_y_address(unsigned int address, unsigned int offset)
{
   int i;
   u32 utemp;
   utemp = (address & SET_LOW_ADD);
    regw_rsz(utemp, offset + RSZ_SDR_Y_BAD_L);
   regw_rsz(utemp, offset + RSZ_SDR_Y_SAD_L);
   utemp = (address & SET_HIGH_ADD) >> 16;
   regw_rsz(utemp, offset + RSZ_SDR_Y_BAD_H);
   regw_rsz(utemp, offset + RSZ_SDR_Y_SAD_H);

    fd_exec(address);
}

The input picture address of the fd_exec function is the output address of the resizer. The fd_exec function call is made in the rsz_set_y_address() function.
I took the 420-Y Output Data of the resizer, because the storing format looks to me like the input format of the FD engine.

Long story short, it doesn't work, so i have some question regarding this subject.

1. Can anybody see any mistakes I made with the FD engine code?
2. Does anybody know the exact input format of the FD engine?
and
3. I have not a clue, how I can handle the Interrupt with the FD engine. So I need help with that!!!

thanks in advance
Alex

Alex,

Recently, I wrote a driver using the EVMDM365 for the Face Detection Module with reasonable success but I abandoned my efforts when I realized that its performance was limited since the image must be monochrome and first written into memory before face detection operations can be performed. This is a serious limitation for our application since we must perform detection on color video at 30 fps but this may not be a limitation for you.

DVSDK 2.10 implemented a fully functional Face Detection Module kernel driver plus it included an example application named face detect_example.c - these items are available on the TI Website.

For our application, we have moved on to an implementation using components from OpenCV.

I hope the above is helpful,

-Chuck