Arduino Deployment Not Visible

Greetings! I’m a teacher learning along side my students as we explore what WIO Terminal and Edge Impulse can do. Kudos on the amazing platform!!!

One of my students created a accelerometer based model using her phone, but does not see the Arduino library option in the deployment tab. Any ideas where she may have gone wrong here? I have sample code once we have an Arduino library, but we don’t have the option to deploy in that format, only C++. Any ideas?

Hi @pway,

Thanks for your great feedback!
Could you share the project ID of your student?

Thanks,
Aurelien

@pway That’d indeed help, as we only ever hide the Arduino deploy option for non-FOMO object detection methods…

The link now does not download anything – it’s a link to an image – is the arduino library still available?

Hi @crippledko,

Can you provide your project ID so we can take a look at your deployment options to see if we can replicate the issue?

I am having the same issue when I trained an object detection model with two classes. There is no option to deploy to an Arduino Library, I am using Arduino Nicla Vision and I have attached a screenshot of the available options to deploy. Here is the project ID : 192284. Any help is appreciated :slight_smile:.

Hello @chinmaysahoo,

We hide the devices that won’t support your impulse.

You won’t be able to run a MobileNetv2 SSD model on the nicla vision, especially on 320x320 images.
You might want to try FOMO models: FOMO: Object detection for constrained devices - Edge Impulse Documentation

Best,

Louis

1 Like

@louis thanks for the reply, if I use FOMO I am getting an option to upload it to Arduino but using FOMO is creating another issue. I am getting a zero F1 score after training. Could you please help me out, and tell me if I am doing anything wrong?

Hello @chinmaysahoo,

I can see that you only have 45 data samples in your project.
I’d start by collecting more data samples.

Best,

Louis

Feature Request
Instead of creating a morphing GUI wherein one must remember the GUI will look like this when I do that, simply overlay an icon or gray out the option with an asterisk * stating the selected device does not support this.

@MMarcial,

We’re currently reworking our Deployment page UI indeed.

Best,

Louis

@louis I trained using FOMO and I am getting a decent accuracy score but after I select the Arduino library in deployment and then build and download the library and try to import it to the Arduino IDE I get this error Error: 13 INTERNAL: Library install failed: archive is not valid: multiple files found in zip file top level, is it because of any IDE version incompatibility or some other issue? Please let me know.

Hello @chinmaysahoo,

You need to delete your old libraries.
If you installed Arduino with the standard method, you can see your Arduino Libraries in User/Documents/Arduino/Libraries. There you can delete the library your previously imported and import the new one from the IDE.

Best,

Louis

Can someone let me know how I can verify whether the memory requirements for my trained ml model is compatible with Arduino Nicla Vision?

Which memory?
The Ardy specs show:


  • Nicla RAM = 1MB
  • Your Model = 2.4M ← Not compatible

  • Nicla Flash = 2MB
  • Your Model = 0.0585MB ← Compatible