Deep Learning NOT for image recognition - no Vision Module

[GIC-SAGM]

Hi.

I am interested/considering the use of AI / deep learning techniques for evaluation of signals from sensors, not for image recognition.
Everything I read relies on Vision Module, which I do not have.

What options are there, if any?

[Andrey_Dmitriev · Accepted solution]

Nowadays most of AI/ML/DL Examples related to DSP are Python scripts, so, this turned your question how to interact between LabVIEW and Python, send your signals from sensors to the Pythons script and receive back the result of the processing.

[GIC-SAGM]

This is of course a complete change of paradigm; good part is that it no longer has much secrets.
I am "fighting back the python invasion", but probably resistance is futile.
We'll be assimilated! 😉

[ngene]

Hi GIC-SAGM,

Please check our DeepLTK (Deep Learning Toolkit for LabVIEW).

Besides CV examples, you can find many non CV ones on GitHub.

For your task, I think, you can consider:

1. Waveform Classifier  - the example is to classify different faults (6 types) on power lines

2. Waveform Regression  - to predict continuous values, e.g. frequency, amplitude, phase, DC offset.

3. Waveform Anomaly Detection - unsupervised learning based approach to identify and localize anomalies on the waveform.

Hope this is helpful.

Please tell me if you need assistance with your development.

[GIC-SAGM]

Thanks for the suggestion.

I have had seen DeepLTK, but to start I'd prefer open source.
For now I am exploring, so my learning curve will surely not fit into the trial period.