PROCESSOR-SDK-AM62A: Complete SDK source code

Hi Xingyun,

Do you have the firmware builder obtained from https://www.ti.com/tool/PROCESSOR-SDK-AM62A#downloads?

Regards,

Jianzhong

Hi, Jianzhou,
We have the Firmware-builder sdk, but we cannot find example codes of AF, could you please give us some hints?

Hi Xingyun,

We don't have example code or reference implementation for auto focus algorithm. We only have that for auto exposure and auto white balance.

Regards,

Jianzhong

Hi, Jianzhou

We want to get the H3A data provided to the AF algorithm, which is not the AF algorithm. Can you give us some hints?

Regards,

Xingyun

Hi, Jianzhou

We want to get the H3A data provided to the AF algorithm, which is not the AF algorithm. Can you give us some hints?

Regards,

Xingyun

Hi, Jianzhou

We want to visualize the AF_H3a data, such as the 0:15 column in the attached file h3a_0000.bin. Are the rules the same as for parsing AEWB_H3a data? Can you give me some hints?

Regards,

Xingyun

1854.h3a_0000.zip

Hi Xingyun,

Please check for my friend request. I'll need to pass you some information under NDA.

Thank you.

Jianzhong

BTW, H3A-AF output data package information is available in "7.5.3.3.4.6.7 RAWFE H3A DMA Interface" of AM62A TRM.

Hi Gang,

According to your suggestion, we visualized the h3a data of AF, but there are some questions to ask.

Is it correct that we use the VAF off model with the same 11 parameter Settings for the two IIR filters, but not the same results for HVF_1_G/R/B and HVF_2_G/R/B? The result of sum_of_pixel doesn't correspond to the original image either, can you give me some hints?

Hi Gang

I still have a confused point on my side. I hope you can give me some tips.

We set the number of paxel to 16*32 (h_count*v_count), there are 1535 (24560/16=1535) lines in the H3A_af.bin file, and all 1535 lines have data. I think this is abnormal, and only the first 512 (16*32) should have data. Please help me to point out where the misunderstanding happened.

Regards,

Xingyun

xingyun zhao said:

Is it correct that we use the VAF off model with the same 11 parameter Settings for the two IIR filters, but not the same results for HVF_1_G/R/B and HVF_2_G/R/B?

The thresholds for HFV_1 and HFV_2 might have been set differently.

Hi xingyun,

xingyun zhao said:

We set the number of paxel to 16*32 (h_count*v_count), there are 1535 (24560/16=1535) lines in the H3A_af.bin file, and all 1535 lines have data. I think this is abnormal, and only the first 512 (16*32) should have data

H3A-AF output should have 16*32*48=24576 bytes.

Hi Gang

The threshold Settings are also the same and are all set to 0

Regards,

Xingyun

xingyun zhao said:

The threshold Settings are also the same and are all set to 0

In this case, HVF_1 and HVF_2 output should be exactly the same.

Hi Gang

What is the maximum window supported by H3A_AF?

Regards,

Xingyun

16H x 32V should be fine.

H/W limitation for HAF-only mode should be 36H x 128V.
I am not sure if there is any S/W limitation in tiovx and PDK driver.

Hi Gang

According to your prompts, each row in the following figure represents the data of a window. Does 06, 24, 04, 00 in the first 32bit in the figure correspond to the first data in H3A-AF? If so, should we read? Whether to read 0x06240400 or 0x042406.

Regards,

Xingyun

Hi Xingyun,

xingyun zhao said:

Does 06, 24, 04, 00 in the first 32bit in the figure correspond to the first data in H3A-AF?

It looks like.

xingyun zhao said:

If so, should we read?

I believe it is in little endian.

xingyun zhao said:

Whether to read 0x06240400 or 0x042406.

should be the latter.