• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 100 subscribers
  • Views 388 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.

EDGE-AI-STUDIO: Anormaly detection on C2000

Ryuuichi machida
Guru 12785 points
Part Number: EDGE-AI-STUDIO


Tool/software:

Hi,

I'm checking following video. According to this video, you described "Simple to train and deploy with on chip learning options for robust anomaly detection".
https://www.ti.com/video/6367452239112

However, I would like you to confirm whether I can perform anomaly detection except "motor fault" or "arc fault" by using edgeai-studio or command line environment.
(For example, I just capture current sensor information from sensor and create dataset on edgeai-studio.)
Is it possible to perform like above ?

BR, 

  • 0
    Guru 12785 points

    Hi,

    Could you please give your feedback about above ?

    BR,

  • 0 in reply to Ryuuichi machida
    TI__Mastermind 20085 points

    Hi Ryuuichi,

    I apologize for the delay on this. I have looped in the Edge-AI expert and they should have a response back soon.

    Best Regards,

    Delaney

  • 0
    TI__Prodigy 360 points

    Hi Ryuuichi,

    The Edge AI Studio Model Composer GUI toolchain can perform Arc Fault Classification/ Motor Bearing Fault Classification.

    As far as you have some time series data collected, and you want to classify them into different faults using the preset options provided by TI, Model Composer should be good enough. So the short answer to your question is yes.

    Apart from that, if you want to explore the full capabilities of AI training on your already collected dataset, you can checkout TinyML Modelmaker for an extensive collection of classification capabilities. This is command line based and needs to be setup manually.

    So in general fault classification is possible on TI's tools, but anomaly detection (I have only normal data, no faulty data) is not yet possible on the tools. It will be supported in a soon enough release later this year.

    Regards,

    Adithya

  • 0 in reply to Adithya Thonse
    Guru 12785 points

    Hello Adithya-san,

    Thank you for your reply.
    >"Simple to train and deploy with on chip learning options for robust anomaly detection".
    For "anomaly detection" in video, I understood that you do NOT have environment to perform this at this time.
    For "simple to train and deploy with on chip learning options" in video, it seems that user can perform not only deploy with inference but also train on C2000 device. Is it correct information ?

    Best Regards,

  • +1 in reply to Ryuuichi machida
    TI__Prodigy 360 points

    Ryuuichi-san,

    As of today, training on device is not supported as a productised publicly available version yet. It should be available before Q4 this year.

  • 0 in reply to Adithya Thonse
    Guru 12785 points

    Hello Adithya-san,

    Understood. I will wait for releasing your onchip training and anomaly detection solution.

    BR,