Compact models on Percept

Question

I have a percept and I found no detail in the documentation but I seem to need a general model from customvision.ai in order to use it on the Percept. There is an error you will get in the Azure Portal if you try to deploy a model that is a General Model but you have the settings set to General Compact. In other words, the settings in customvision.ai are out of sync with the most recent iteration of trained models in the project. The error in the portal isn't helpful at all. It is a bit strange that only full general models are available on Percept as you pretty much need to use compact to export them anywhere else. Will this support compact models in the future? What is the reason for this requirement? Why isn't it documented anywhere?

Answer 1

@Ben Kotvis - Azure Percept only supports compact, compact S1, and general compact models from Custom Vision. Currently when building a model with Custom Vision it is recommended to do so within Azure Percept Studio rather than directly in customvision.ai. That way you can ensure it will work on the Azure Percept DK. This will be made more clear in the documentation (thanks for pointing it out).

With regards to the error message, please provide the error text so it can be addressed. Thanks!