Could I upload image to invoke an image recognize application?

[archichen]

[Valenzione]

Hi, since it's not an issue I'll convert this to a new discussion 😉

[archichen]

wow, cool, this may new way to discuss on github

[tidylobster]

HI @archichen. You can submit images as tensors to your models. The conversion to tensor is typically done on the client's side and the resulting tensor is sent for prediction.

[archichen]

Thank you for reply! But I'm not an AI engineer and have not seen related content on the Hydrosphere website... So that I do not understand that how to convert image to tensor and send to model. Could you provide document or example? Such as how to package images and how to invoke.

[tidylobster]

At the moment there aren't' any tutorials on image processing in the documentation. But we have a few examples of processing MNIST images, although they are outdated. I'll make a quick draft with the updated examples and share it here.

[tidylobster]

Here, you can check this draft for the latest working image example: Hydrospheredata/hydro-serving-example#51. If you have any questions, I'm happy to help.