When you upload a document that contains tables, Nanonets AI automatically extracts the table data as rows and columns. If the table was not detected, you may need to train the AI to capture data from your documents accurately. Watch the video below to learn how to capture tables from scratch.
How to capture custom columns
A. Add custom Column names to your model
-
Visit the “Training Files” section of your model and upload your test documents. (At least 10 unique files with the same columns). The AI will use these files to learn.
-
Add the names of the columns you want to capture (eg: Credit, Debit). Do do this, click on "Add table header" under the Tables section
-
Type the name of the column and then click on "Add table header" to confirm. You can also add more columns while labelling the files for training.
B. Mark examples for the AI to learn
- Click on one of the images you uploaded.
-
If the table was already captured, simply select the respective column names from the dropdown for each column you need.
- Do this till each column has at least 10 examples marked. You can find your progress here:
- Click on Train Model once 10 examples per column are marked.
C. If the Table was not auto-captured, you can draw a table from scratch. Follow the steps below.
- Click and drag across the table on your image
-
On the label pop-up, switch to Table.
-
Click on Add Table to place a 1 x 1 grid. (You can adjust the rows and columns once the table is placed)
-
Hover over the edges of the table grid. A blue circle will appear. Click to add a column separator at that point.
-
Similarly, let’s add row separators.
-
You can reposition the line by dragging it if you make a mistake.
-
You can delete a line by clicking on the red dot, which appears on the edge of a line.
-
Once done, click on “Extract data as table.”
-
This is how the capture would look like 👇🏽.