We're excited to assist you in evaluating the platform once you're on the Pilot with the Pro Plan.
To make this a success for you, here are the next steps:
1. Introducing you to our Onboarding Team:
2. Log In
Log into Nanonets App with your email address
3. Select Model Type
You can select a Pretrained model type - like an Invoice model for invoices. These models are pretrained with millions of files to capture data instantly.
If your document type is not present or you have a unique requirement, you can create a Custom model.
Using a Pre-Built model
If you have selected any document types: invoices, tables, receipts/bills, US driver's licenses, passports, or purchase orders, then follow the steps below or learn how to train a custom model.
-
Step 1: Upload your documents on the Extract Data screen
-
Step 2: Verify your data by clicking on the file
-
Step 3: Download data by clicking on the button at the bottom left
Voila! You can now extract data using Nanonets like a pro! Wasn't that a piece of cake? :
Using a Custom model
If you are building a custom model or trying to improve the accuracy of your current model, you will have to train the Nanonets AI to capture data from your documents. You can watch this video or follow the below steps.
a. Please upload at least 10 sample images or PDFs of the document type you want to process. You can add them to the Upload screen of your model ID or automate them through Google Drive / Dropbox.
Tip 1: More the number of images you upload, the better the model accuracy
Tip 2: We suggest that the data be as diverse as possible (different layouts, formats, templates, vendors, etc.) This ensures that the model is accurate and can predict well on any new template, even if it has never seen that one before as you test the platform or use it in production in the future.
b. Please update all the labels you want to capture from the document type. You can add them on Manage Label screen on the model.
c. Please annotate at least 10 samples of the files with the labels you want to capture. You can label the samples on the platform on the "Show Labels" screen.
Training the Model
It takes ~1 hour to train the model, post which you can test it out.
Testing the Model
After the model is delivered, your team can test the model and integrate the API for the next 2 weeks. Please share your feedback so our team can address those concerns in the best way possible.
Please reach out to us if you have any questions.