Training two data sets : Sensor fusion

Question/Issue:

Project ID:

Context/Use case:

Hello Edge impulse, I am trying to train two data sets (audio and images) for specifically beehives in one model. Because i want to measure accuracy angainst power later and see their performance. But I cannot go further at impulse design in having two axes and two inputs for these two kinds of data. I have only managed to train these individually.

Secondly I am to train both temperature and humidity in one. But accuracy is too low. Please assist me on this.

Hi @ephrance,

You cannot use both audio and image data types in a single project. Your only way of doing that would be to convert both to a series of raw data (e.g. arrays) and upload those to your project (e.g. in CSV or CBOR format).

For the second problem, you have a number of things to try:

  • Gather more data
  • Increase epochs
  • Try different model architectures
  • Try the EON Tuner to see if a particular combination of processing and ML blocks work well together

Without insights into your data, I cannot help much. Please provide your project ID if you would like one of us to look at your data.