Question: How to Input Images Directly into WebAssembly Model?

Help
#1

Hello everyone,

I have successfully run my trained FOMO model as WebAssembly on Windows. I already have the images that I want to use for object recognition. However, the WebAssembly documentation only mentions using raw features from Edge Impulse’s Live Classification.

Is there a way to input images directly into the WebAssembly model?

Thank you!

#2

Hi @Hisamichi.Harada

Yes that is possible, you can pass an image as you do for other examples, see the classification via JS example for running via NodeJS locally or the Web Browser example

Best

Eoin

#3

Hi @Eoin

Thank you for the advice.
It seems that in the deployed WebAssembly, we need to input ‘Raw features’ from the image. I understand that I can upload an image to Data Acquisition to obtain the Raw features, but is there a way to do this locally?
I’m using images generated from simulations, so I won’t be doing real-time object detection with a webcam.

Thanks again for your help!

#4

Hi @Hisamichi.Harada

You should be able to use Canvas getImageData and pass that instead:

Best

Eoin