• State Resolved
  • Locked Locked
  • Replies 9 replies
  • Subscribers 98 subscribers
  • Views 460 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TDA4VM: Fvid2 API call timing confirmation

yuanjia cai
Prodigy 50 points

Part Number: TDA4VM

Requirement：

   1.We have a number of different sensors,channel independent.

   2.The image output by each sensor needs to be processed by the VISS.

Questions:

   1. Since there is currently only one VISS, how to handle the difference in VISS configuration for each image channel?

   2. What is the order in which FVID2 is called?

Our try:

   1.To handle channel differences, we try to call FVID2_SET_CONFIG before each graph is processed.

   2.Following the above process, there is a high probability that the VISS FVID2_processRequest return -1 and the FVID2_getProcessedRequest return -4 will occur.

  • 0
    TI__Guru**** 494965 points

    Hi,

    Please find answers to your questions below. 

    yuanjia cai said:
     1. Since there is currently only one VISS, how to handle the difference in VISS configuration for each image channel?

    The request for VISS processing needs to be serialized in this case. Driver supports this feature. If you open multiple handles of the driver, and submit multiple requests to the driver for each of these handles, driver internally submits the requests one by one. 

    yuanjia cai said:
    2. What is the order in which FVID2 is called?

    Which release are you using? In earlier release, heterogeneous cameras is not supports. If this is supported, then you could open the driver, configure it and then submit the frames for each handles. Driver will internally switch the configuration..

    yuanjia cai said:
       2.Following the above process, there is a high probability that the VISS FVID2_processRequest return -1 and the FVID2_getProcessedRequest return -4 will occur.

    Can you please provide some more information on when processRequest returns -1 and when getProcessRequest returns -4? For which handles? What is handle configuration?

    Regards,

    Brijesh

  • 0 in reply to Brijesh Jadav
    Prodigy 50 points
    Brijesh Jadav said:
    The request for VISS processing needs to be serialized in this case. Driver supports this feature. If you open multiple handles of the driver, and submit multiple requests to the driver for each of these handles, driver internally submits the requests one by one. 

    We will process each image serially in a task. 

    Brijesh Jadav said:
    Which release are you using? In earlier release, heterogeneous cameras is not supports. If this is supported, then you could open the driver, configure it and then submit the frames for each handles. Driver will internally switch the configuration.

    We develop based on SDK_7_3.

    Does the heterogeneous camera support in SDK73?

    Brijesh Jadav said:
    Can you please provide some more information on when processRequest returns -1 and when getProcessRequest returns -4? For which handles? What is handle configuration?

    We call AppViss_SetAllConfig every frame with the same parameters, the current frame rate is 20Hz per channel, the same trigger source triggers 3 cameras, and each image uses VISS VHWA_M2M_VISS_INPUT_IDX0.

    If you want to reproduce the relevant problem points, you can try to use the above methods.

  • +1 in reply to yuanjia cai
    TI__Guru**** 494965 points

    Hi yuanjia cai,

    yuanjia cai said:

    We develop based on SDK_7_3.

    Does the heterogeneous camera support in SDK73?

    No SDK7.3 does not support heterogeneous camera. I think it was added in the SDK8.0 release. There is an init time parameter in the viss init that enables this feature in SDK8.0 release.

    yuanjia cai said:
    We call AppViss_SetAllConfig every frame with the same parameters, the current frame rate is 20Hz per channel, the same trigger source triggers 3 cameras, and each image uses VISS VHWA_M2M_VISS_INPUT_IDX0.

    Are you using test0 in gAppVissTestConfig array from the file ti-processor-sdk-rtos-j721e-evm-08_00_00_12\pdk_jacinto_08_00_00_37\packages\ti\drv\vhwa\examples\vhwa_viss_test\vhwa_viss_cfg.h? This is because test0 is 720p resolution, whereas it seems you are using 1080p resolution. Have you changed anything in this test configuration? Is it possible to share updated file & test case? 

    Regards,

    Brijesh

  • 0 in reply to Brijesh Jadav
    Prodigy 50 points
    Brijesh Jadav said:
    No SDK7.3 does not support heterogeneous camera. I think it was added in the SDK8.0 release. There is an init time parameter in the viss init that enables this feature in SDK8.0 release.

    Is there any way to verify the functions of heterogeneous cameras on 8.0 based on version 7.3?

    If yes, can you provide a patch? If not, can you explain the reason?

  • 0 in reply to yuanjia cai
    TI__Guru**** 494965 points

    Hi yuanjia cai,

    Do you mean to support Heterogeneous camera on 7.3 release? This is not supported in SDK7.3 release. You would have to backport the changes from 8.0 to 7.3 release. The changes are mainly in the PDK VHWA drivers.

    Regards,

    Brijesh

  • 0 in reply to Brijesh Jadav
    Prodigy 50 points

    Our team is now developing based on 7.3. Show no intention to migrate to 8.0.

    Is it possible for me to use a patch to verify the function of heterogeneous cameras?

  • 0 in reply to yuanjia cai
    TI__Guru**** 494965 points

    Hi Yuanjia cai,

    We dont have support or patch to verify this feature on SDK7.3 release. As i said, you would just require to change vhwa drivers, especially viss driver to have this feature in SDK7.3. 

    Regards,

    Brijesh

  • 0 in reply to Brijesh Jadav
    Prodigy 50 points

    Which demo can verify heterogeneous camera configurations? I now have an EVM environment that can be used to run verification.

  • 0 in reply to yuanjia cai
    TI__Guru**** 494965 points

    You could run single-camera and multi-camera example with different input sensor. That's one way to verify this feature.

    Regards,

    Brijesh