• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Answers 2 answers
  • Subscribers 98 subscribers
  • Views 456 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:

TDA2EVM5777: Caffe Jacinto Model failed to train SSD

TDA2EVM5777: Inference using the trained model with caffe-jacinto

LY_Liew
Prodigy 190 points

Part Number: TDA2EVM5777

Hi,

I am trying to run this command "python ./infer_video_object.py" from the guide here.

No error message was shown after the execution of this command.

However, when I looked into the output folder defined by "params.OpPath", the output video only had a frame in it without any inference, with a clean console_err.log.

Please help on this issue.

Thanks.

Best regards,

Lityang

  • 0
    TI__Genius 17580 points

    Hi Lityang,

    I have forwarded your query to respective team and awaiting for their response.

    Thanks,

    Praveen

  • 0 in reply to Praveen Eppa1
    TI__Genius 11546 points

    Hi Lityang,

    This might have happened due to a recent unintentional change in caffe-jacinto-models repository. I have pushed a change that is expected to correct the issue.

    The git history has changed, so if the git pull fails, you can clone the repository again.

    Could you please clone the caffe-jacinto-models repository and check again?

    Best regards,

    Manu.

  • 0 in reply to Manu Mathew
    Prodigy 190 points

    Hi Manu,

    I tried this version that you pushed recently, but some bugs were still found. I will list down the bugs found below:

    1. No propagate_obj.py required by ssd_detect_video.py was found in this latest repo.

    2. Inconsistent parameters in ssd_detect_video(params) in both infer_video_object.py and ssd_detect_video.py (enObjTracker).

    3. Syntax error at line 235/236 in ssd_detect_video.py.

    Nevertheless, I was finally able to run the inference after I modified the codes.

    By the way, for your information, I found that the previous version of the said repo could actually work once I changed the params.NumFrames to a larger integer (was originally 1), this might be the problem that caused the one-frame-only output.

    Thanks for the helps.

    Best regards,

    Lityang

  • 0 in reply to LY_Liew
    TI__Genius 11546 points

    Hi,

    Thank you for the information. We have pushed a commit to fix these errors.

    Best regards,

    Manu.