TDA4VM: support Vision Pre-processing Accelerator (VPAC) on the linux

cesar.xu

Intellectual 470 points

Part Number: TDA4VM

Hi,

Our camera is raw data image, which needs to be converted to yuv image through vpac module,

please provide corresponding support in linux.

Regards,

Cesar

over 2 years ago

0 Brijesh Jadav over 2 years ago

TI__Guru**** 496655 points

VPAC is supported only on R5F, you can use VPAC OpenVX nodes to access VPAC functionality from Linux.

Regards,

Brijesh

0 cesar.xu over 2 years ago in reply to Brijesh Jadav

Intellectual 470 points

Hi Brijesh,

Please provide relevant documentation and routine tests, thank you.

Regards,

Cesar

0 Brijesh Jadav over 2 years ago in reply to cesar.xu

TI__Guru**** 496655 points

Hi Cesar,

We have single camera and multi-camera demo in vision apps, which uses VPAC functionality from R5F. Could you please refer to these demos?

Vision Apps User Guide: Single Camera VPAC Application

Vision Apps User Guide: Multi Camera VPAC Application

Regards,

Brijesh

0 cesar.xu over 2 years ago in reply to Brijesh Jadav

Intellectual 470 points

Hi Brijesh,

Csi2rx gets the camera data from the Linux end, which is /dev/videox;

and get camera raw image on the linux, how use vpac.

Regards,

Cesar

0 Brijesh Jadav over 2 years ago in reply to cesar.xu

TI__Guru**** 496655 points

Hi Cesar,

Maybe you can use gStreamer plugin to process the data on Linux. There is a gStreamer plugin available for VISS. But i think it is part of edgeAIsdk. If you can use it, i will try to get more information and share documentation.

Regards,

Brijesh

0 cesar.xu over 2 years ago in reply to Brijesh Jadav

Intellectual 470 points

Hi Brijesh,

Brijesh said: “ i will try to get more information and share documentation. "

ok, you can provide relevant documentation.

Regards,

Cesar

0 Brijesh Jadav over 2 years ago in reply to cesar.xu

TI__Guru**** 496655 points

Hi Cesar,

Please refer to below documentation,

1. Overview — Processor SDK Linux for AM68A Documentation

https://software-dl.ti.com/jacinto7/esd/processor-sdk-linux-am68a/09_01_00/exports/edgeai-docs/common/edgeai_dataflows.html#object-detection

https://software-dl.ti.com/jacinto7/esd/processor-sdk-linux-am68a/09_01_00/exports/edgeai-docs/common/configuration_file.html#pub-edgeai-configuration

Regards,

Brijesh

0 cesar.xu over 2 years ago in reply to Brijesh Jadav

Intellectual 470 points

Hi Brijesh,

How do the following files compile and generate themselves?

Regards,

Cesar

0 Brijesh Jadav over 2 years ago in reply to cesar.xu

TI__Guru**** 496655 points

Hi Cesar,

I know from the RTOS side. In RTOS size, under imaging folder, each sensor has its own generate_dcc.sh script. Running this script generates this dcc binary files.

Regards,

Brijesh