Detecting Model with Microsoft Azure Cognitive Services: Form Recognizer

Question

I have created a custom model that works well with the Azure Form Recognizer.

I plan on using multiple models including the inbuilt models. I would like to be able to detect the best suited model (for example if I scan a receipt, it should use the inbuilt receipt model, or if I scan in a custom form it should choose the best fit model from the multiple that I have trained)

I am currently using (but not restricted to) the C# SDK, I am currently using this code:

            RecognizedFormCollection forms = await recognizerClient
            .StartRecognizeCustomForms(modelId, ms)
            .WaitForCompletionAsync();

Notice that I StartRecognizeCustomForms with a modelId, so I have to provide it a particular modelid for each document..

Edit I have discovered I can create "Composed Model's" - however I cant see any way to include inbuilt (receipts ect) models into these compositions.

Answer 1

Using this Form Recognizer Labeling tool makes it easier to extract the data based on the layout. if you have multiple variety of forms you can use the Composer option on the v2.1.1(still in preview) to merge multiple models and run the forms against the composed to model to automatically extract the data.

Please follow the below samples for recognize receipts. https://learn.microsoft.com/en-us/dotnet/api/overview/azure/ai.formrecognizer-readme-pre#recognize-receipts

Currently it's not possible We are checking with the team for roadmap to include the inbuilt receipt model.