Can I use Edge Impulse for a different purpose than intended?

I want to build a model that classifies images into categories. its input is a single image, and the output is the class to which the model thinks the image belongs.
But it seems like the model searches for a camera feed as input.
How can I use the model the way I need, with a single image input instead of live camera input? is there a build-in option for this kind of use?

thank you,


@liorkot you can upload data from your connected device or any other device which contains your image using web uploader or ingestion service API

Can I do it in the Live Classification part or when I run it on my computer?

@liorkot your questions is not clear at this point, are you trying to classify images on basis of model trained in your edgeimpulse studio?

I’ll explain in more details -
I’m trying to build a model that receives an image and classifies it into one of two categories.
the model I built with edge impulse is working well on the test data, with 95% accuracy which is more than enough for the purpose i need the model for.
but the problem is that i need the model to receive a single image as input, and to output the class of the image. that’s problematic because it seems like the model built by edge impulse can only receive a camera live feed as input.

my question is - is there a way to make the model receive a single picture as input?

Yes you can do this task easily by using Ingestion Service API and if you want to run your model on your local machine you can use the code from the model as python notebook or just use API for this task.

Hello @liorkot

If you’re on MacOS or Linux, you can use Edge Impulse for Linux to run it locally.
Here is some examples using the Python SDK that will let you classify single images (or even a video frame

Feel free to have a look at the documentation too:

If you want to use the Live Classification, you can upload a new image in your Test Dataset through the Uploader or using the ingestion service API as @AnshumanFauzdar mentionned.



1 Like