Inconsistent results in WASM image classification

[Gigcity]

Project ID:372531

Following example documentation for webassembly, I’m looking to pass an image url (JPG) for classification based on a model. The results of 2 features vary across different methods.
The example only shows testing via copy/paste of raw features to compare in console output, which seems to match.
Without any existing examples, I’ve attempted to pass the same image for classification a hosted WASM. One test via server.py in Windows, and the other on a web server. I would have expected the later two to be identical, but they are not, and neither match the raw feature test.
I’m out of ideas - see attachments.

[Eoin]

Hi @Gigcity

Browsers and architectures possibly it’s hard to tell what part of your custom code / buffer array processing could be causing differences. Try debugging or changing how you work with the JPG in a simple use case first.

Perhaps check around for WASM discussion forums to get an opinion on browser / standalone differences here. Let us know what you find!

Best

Eoin