We are excited to release a new public cloud endpoint for Document Understanding Machine Learning Classifier Model. Our vision is to create an end-to-end document processing platform that can cater to any and every document type and all varied use cases. Classification is usually the first step in a document processing process. With this pre-trained document classifier public endpoint, customers can easily leverage the most state-of-the-art machine learning without having to annotating documents to train the model. The model is pre-trained for 10 document types including receipts, invoices, purchase_orders, utility_bills, passports, ID_cards,RemittanceAdvices, DeliveryNotes, W2 and W9.
How to use the Machine Learning Classifier endpoint?
If you have never used UiPath Document Understanding product, please review our documentation:
The classifier endpoint works with the Machine Learning Classifier activity. Drag and drop the Machine Learning Classifier activity to the Classify document scope as below.
Input the public endpoint https://du.uipath.com/classify/MLclassification and the Document Understanding ApiKey.
Configure Classifiers and match the document types. Now it is ready to run!
Hi @Monica_Secelean ,
Exciting new addition to the DU suite!
Do you know what’s the pricing for consumption of this endpoint?
@warren_lee happy you like it - find pricing details here: Metering & Charging Logic
Hi UiPath team. This is a really cool feature!
I tried it out with one file having multiple document types (invoices, purchase_orders, receipts, utility_bills, etc) and the answer was one unique classification result. Should not it generate at least one classification result for each document type?
Hey @felipe.restrepo , thanks for giving our new feature a try
The classifier is able to classify one document type only - but stay tuned, we’re planning on adding a splitter which would be able to split your document, so that you can then classify each one
Keep the feedback coming & thank you!
@Monica_Secelean, Thanks for the heads up!
Do you have any idea about the roadmap/timeline for this split feature?
In the mean time and out of curiosity, I’ll try splitting and classifying each page, and later merging all the classification results to decide latter how to split the whole document
@felipe.restrepo If everything goes well, we’ll have something for 22.10