50
Hi, anyone know if there are technical limitations to custom document processing models?
We have a custom model trained to get a table across 4 pages of an invoice. The training set of documents are able to detect it fine during the tagging and testing inside the model dialogue.
However, on the actual usage of the model, it always emits the 4th page consistently on the same documents used to train it.
Anyone encountered this before?
However when you are using your model in a flow , it omits the 4th page ?
---Correct
Did you specify a page range to process ?
---Yes, current range is 3-6, but page 6 is omitted.
---If I specify just page 6 (the specific page thats being omitted), it works ok.
Feel free to send us an email to aihelpen@microsoft.com with your samples so we can investigate better?
---Done, thanks!
Just to make sure I understand.
You were able to tag all tables correctly on all your documents and train a custom model.
However when you are using your model in a flow , it omits the 4th page ?
Did you specify a page range to process ?
Feel free to send us an email to aihelpen@microsoft.com with your samples so we can investigate better?
Thank you!
Do you have the same behavior for all your documents that you use in this custom model?
--Yes, worked fine via the training and testing from inside the custom AI model menu.
Do you use this model in a flow already?
--Yes, the flow only detects 4 pages.
Could you please share screenshots?
--This is the training data:
There is no limitation
Do you have the same behavior for all your documents that you use in this custom model?
Do you use this model in a flow already?
Could you please share screenshots?
thank you!
Share
Report
All responses (
Answers (
