Get our latest book on "Top 10 artificial intelligence myths."

Last updated

March 3, 2021

Table of contents

How to retrain your model

This tutorial will walk you through how to retrain your model. By the end of this tutorial, you will learn how to quickly and easily:

  • Confirm a prediction
  • Edit a prediction
  • Delete a prediction

These three simple interactions provide immediate feedback to Impira AutoML to improve the performance of your model and see updated results in real time.

Let’s get started.

Impira AutoML predictions are visually represented in the User Interface (UI) using colors and flags: Green indicates a highly confident result, red suggests that a review is recommended, and black denotes a field you have confirmed or added manually. At a glance, you can easily ascertain areas that need a review, or others that have higher quality results. 

Double-click on the cell you want to confirm. Impira will take you to the exact location of the ML prediction in your File. The field you double-clicked will automatically be highlighted in the right-hand sidebar. 

For each of the extracted fields you can confirm the prediction, edit the OCR text prior to confirming prediction, modify the bounding box or remove the predicted value and start over.

Confirm a prediction

Click on the “tick” button next to the extracted value.

Edit the OCR text prior to confirming the prediction

Click on the field name or extracted value to open the extended field view. From here you can also make edits to the extracted OCR text before confirming the prediction. 

Modify the bounding box

Click on the field in the right-hand sidebar, or click on the colored bounding box for the field you want to update. Click and drag the corners of the bounding box to the appropriate dimensions. You can also click and hold your cursor on the box to move it over the top of the data you want to extract. Once you are happy with the placement and extracted data, click on “Confirm value”.

Remove the prediction and add a new field

In rare cases, you may find it easier to remove a field and add it again. Do this by clicking on the field in the right-hand sidebar and clicking the “X” in the upper right corner of the edit field box. This will remove both the ML prediction and extracted data for the field, and you can now add it again making necessary refinements to your selection.

And that’s all there is to it. You now know how to retrain a model, giving you higher quality predictions faster than before.