• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 98 subscribers
  • Views 451 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.

Original question:

TDA2PXEVM: TDA2 : About Tidl_OD usecase

TDA2SX: Multiple queue with single channel to single queue with single channel output

SagarK
Intellectual 705 points

Part Number: TDA2SX

Hi, 

I am referring to tidl_OD usecase. and I have designed my usecase on similar lines as tidl_OD. Only difference is tidl_OD has grpx which overlays output on dupped input video.

In my case, I have alg link which draws on video frame itself and pass it to Display.

eg. Camera -> Alg_preProc(A15) -> Alg_Split(IPU)

Alg_Split(IPU) -> Alg_tidl(EVE1) -> Merge

Alg_Split(IPU) -> Alg_tidl(EVE1) -> Merge

Alg_Split(IPU) -> Alg_tidl(EVE1) -> Merge

Merge ->Alg_postProc(A15) -> Display

After merge, I am getting single queue with 3 channels output. But, my Alg_postProc is expecting single channel input and finally display also expects single channel input,

Can you comment , how can I get single queue single channel output after merge ?

 

Regards,

Sagar

  • 0
    TI__Genius 14305 points

    Hi,

    Can you explain the functionality of the custom algorithm used in the usecase?

    How are you creating the usecase? Are you using the tool?

    Are able to see image on the display?

    Thanks

    Gaviraju

  • 0 in reply to Gaviraju Doddabettahalli Bettegowda
    Intellectual 705 points

    Hi Gaviraju,

    1. Custom algo links are basic only 

    Alg_preProc = It is mainly for resizing purpose to make input tidl compatible (eg. interleaved to planar)

    Alg_tidl = It is similar to ssd based object detection which is present in vision sdk demo. (some changes we have done and we are processing everything on eve)

    Alg_postProc = It is to draw bounding boxes (using opencv like libraries ) received from tidl output on input buffer itself and then pass it to display

    2. Yes I am using tool present in Vision SDK

    3. Following usecase is working fine and I can see output on display with bounding boxes as well :-)

    Camera -> Alg_preProc(A15) ->  Alg_tidl(EVE1) -> Alg_postProc(A15) -> Display

    But, above design is providing only 10fps. Because, tidl is taking higher execution time. So, after referring tidl_OD use case, I understood that if we use multiple eves then we can achieve higher fps. So, I followed the same design. For my design, You can think of tidl_OD without dup of input capture and grpx overlay.

  • 0 in reply to SagarK
    TI__Genius 14305 points

    Hi,

    Use the Sync link after the Merge link.

    Thanks

    Gaviraju