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: Need Clarification on initial YUV frames getting garbage(FF frames)

Part Number: TDA4VM

While capturing the yuv frames through customised app_single_cam test app, we are getting 0xFF data in first 6-7 frames.
How to resolve this issue.

Thanks

  • Hi Anant,

    Could you give more information on your setup, 

    Which sdk, camera, configuration and what changes in the demo ? and how are you viewing the frames ?

    Regards,
    Gokul

  • Hi Gokul,

    SDK : version 10.1

    Camera- OV camera

    Configuration- GMSL2 + app_single_cam.cfg

    No changes in demo 

    Viewing frames using 7yuv app

    Thanks.

  • Hi Anant,

    Could you verify if the camera is giving the correct image frames from the starting. 

    Viewing frames using 7yuv app

    So you are saving the output directly from the capture node ?

    Could you check the capture statistics to check if there are any errors reported by csirx ?

    Regards,
    Gokul

  • Hi Gokul,
    Yes camera is giving correct frames from starting.

    No, I am not saving output directly from capture node, capturing YUV output from ISP (VISS node)

    Checked the capture statistics, there are no errors reported by csrix.

    Regards,
    Anant.

  • Hi Anant,

    Yes camera is giving correct frames from starting.

    How are you checking that ?

    Camera- OV camera

    Could you describe your entire pipeline, and more info on the sensor, I thought the sensor gives yuv frames as output.

    is this sensor driver supported out of box or you have written your own driver.

    Could you dump the raw frames from the camera and view it externally to isolate its a csirx issue or vpac issue.

    Regards,
    Gokul

  • Hi Gokul,


      We are following the below steps 
      1.We start the sensor
      2.Immediately prime the pipeline
      3.Then dequeue and use frames

    We currently do not have explicit validation to confirm that frames are correct from the very first frame.
    We assume correctness based on:

    • No timeout errors during dequeue
    • No error frames detected

    The capture node receives RAW frames, which are then passed through the VISS (ISP) node to generate YUV (NV12) output. The saved frames are taken after ISP processing, not directly from the sensor or capture node.

    Current pipeline

    Sensor (OV camera) -> CSI Capture (tivxCaptureNode) ->RAW buffers (tivx_raw_image) ->VISS (ISP node) ->YUV (NV12) ->saveImage() → .yuv files
    We are not saving output directly from the capture node.
    We are saving post-processed YUV frames from the ISP.
    Yes, we are able to dump the raw frames and view it through YUV app.
    Thanks and Regards
    Anant

  • Hi Anant,

    Could you dump the RAW image from capture node output and view that file offline using any raw image to rgb or yuv converter.

    This can help isolate the issue from the capture and viss part. And have you done your isp tuning ?

    is this sensor driver supported out of box or you have written your own driver.

    ?, are you using ov2312 sensor driver which is there in the sdk ?

    Regards,
    Gokul