Uploading and training on already processed data

Question/Issue:

I have an already existing sensor on which I have been collecting process data for a long time. The process data contains the features I would like to train a classifier on(RMS,…). But I can’t figure out how I use processed data to train a model. I already tried using the raw input, but it generates a feature for each data point.

What would be the correct way to use preprocessed data in edge Impuls?

Project ID: 174943

Context/Use case:

Train model on existing process data.

@Thijs welcome to the EI community.

By pre-processed data I assume you mean you want to add additional Features like summary stats for each Sample such as:

  • Mean
  • Median
  • StDev
  • Skew
  • Kurtosis

To each Sample you can add additional data before uploading to EI Studio. Add a comma and then the new data. See the protocol doc.

You can also do this inside the Edge Impulse Studio via Custom Processing Blocks.

See this tutorial.

Also watch this video which has some good info on cppType.

Other useful related links:

Forum post about cppType

C++ Library

Running on a Desktop (handy for troubleshooting Custom Blocks))

1 Like