AM5749: TIDL predictive maintenance demo

matusan

Part Number: AM5749
Other Parts Discussed in Thread: TMDSIDK574

Hello, TI Experts,

We have a question about GUI-demo(TI Deep Learning) in PROCESSOR-SDK-LINUX-AM57X like below.

http://software-dl.ti.com/processor-sdk-linux/esd/docs/06_02_00_81/linux/Examples_and_Demos/Application_Demos/PdM_Anomaly_Detection.html#deliverables-from-processor-sdk-linux

We success to run "Predictive Maintenance Demo" on TMDSIDK574.

And we are investigating how to create "onnx model" of this demo by following below TI guided link.

https://git.ti.com/cgit/apps/pdm-anomaly-detection

We also found the file "timotor20190227raw.onnx" provided by TI(Hongmei Gou-san).

https://git.ti.com/cgit/apps/pdm-anomaly-detection/tree/models/timotor20190227raw.onnx

Question:

- Could you tell us the recommended way of creating "timotor20190227raw.onnx"?

We found the below comment in "timotor20190227raw.onnx".

"MATLAB.Deep.Learning.Toolbox.Converter.for.ONNX.Model.Format"

So, we would appreciate if you provide RNN/LSTM model training-script/environment for generating "timotor20190227raw.onnx" of MATLAB.

Best regards,

over 4 years ago

0 matusan over 4 years ago

Guru 10235 points

Hi,

Thank you for your help.

Do you have any update?

Now, we are waiting for the answer.

I would appreciate if you tell us about the progress on this.

Best regards,

0 Michael Erdahl over 4 years ago in reply to matusan

TI__Genius 13455 points

Hi,

Apologies this was missed, we are looking at this now.

Regards,
Mike

0 matusan over 4 years ago in reply to Michael Erdahl

Guru 10235 points

Hi,

Thank you very much for your kindness.

I really appreciate your help.

We are waiting for the answer.

Best regards,

0 matusan over 4 years ago in reply to matusan

Guru 10235 points

Hi,

Thank you for your help.

Do you have any update?

Best regards,

0 Manisha Agrawal over 4 years ago in reply to matusan

TI__Mastermind 22386 points

Hi Matusan,

Offline training in Matlab is applied to get the ONNX model:

The Matlab code is not available publicly. Support for predictive maintenance use case is limited at the moment.

Regards,

Manisha

0 matusan over 4 years ago in reply to Manisha Agrawal

Guru 10235 points

Hi,

Thank you very much for your kindness.

I really appreciate your help.

I understand the situation.

We would appreciate if you consider to share the "predictive maintenance demo" training environment such as caffe-jacinto in future SDK release.

Best regards,

Processors

Processors forum

AM5749: TIDL predictive maintenance demo