Set up a form with a master image
An account on FormX portal
An understanding on background of FormX's basics
Sample business/name cards that will be used in this tutorial
Obtain the following information from the business/name cards: {name}, {job title}, {email}.
Navigate to the form listing page of FormX and click the "Add New Form" button on the top right corner.
You will be prompted with a Create Form modal. Name this form "name card demo" and upload the master_image.jpg from the downloaded sample name cards. Then click "Create".
A new form will be created. Click on the "name card demo" new form.
You will be redirected to the form editor page.Now, you can start setting it up so that FormX can extract data on the sample name cards based on this form.
Now, with the newly created form, we'll go through a few concepts before marking areas on the master image.
An Anchor is a labeled area that serves as a positional reference point, allowing FormX to warp uploaded images of this form. Since uploaded images are more or less different (i.e. a tilted angle) from the master image, FormX will twist them by matching their Anchors to the master form's ones before extracting.
For every form with a master image, a user has to mark at least two Anchors, with the longest distance possible between them. When they are further part, more of the master form resides between them, which will improve extraction accuracy since the warping process becomes more consistent.
TL;DR: Like an actual anchor, drop one on your master form to prevent future uploaded forms from moving away. At least two are required for FormX to function properly.
You may be wondering: Which parts in the master form should I include with an Anchor? Identify the common parts of your documents (with the same format, of course), then include them with an Anchor.
In the example, we'll mark four anchors — name, telephone number, Hong Kong (HK) and Taiwan (TW) address. Name and TW address are especially important, as they are respectively the topmost and bottom-most element in our name cards. Anchoring them significantly facilitates the warping process.
Choose the second tool as shown in the image below to mark an Anchor.
Then mark four points as shown in the GIF below. Note that to create an Anchor, the final point must connect with the first.
Once the name and telephone number are marked, scroll down and include the two addresses as well.
The master form now has enough Anchors, which can then be used to check the corresponding parts of future uploaded name cards.
Detection Regions are the areas that are marked from which you need data extracted. One Detection Region can have several extraction fields, as one region can contain several items of data that are worth obtaining or extracting.
TL;DR: Mark the area where you want information extracted from on the master form with Detection Region.
In this example, we'll extract name, job title and email for these name cards. To do so, they will be marked via Detection Regions.
Choose the third tool marked with a red box in the screenshot below:
As shown below, we'll add three Detection Regions, each containing a field called ‘name,’ ‘job title,’ and ‘email’.
Click the "Save" button. Now you've finished setting up this form!
There are two more name cards in the ZIP file you've downloaded at the start of this tutorial. They will be fed to our freshly created form to have the data targets extracted.
Navigate to the "Test" tab and choose sample1.JPG. This will trigger an extraction which will complete in no time. Repeat these steps with sample2.JPG and you will get similar results.
In both cases, you can see that the input images are warped and "aligned" to the master form. The extraction results are listed out with each image's corresponding fragment, and enclosed with a JSON value at the very bottom.
Images taken my mobile devices (i.e., smartphones) work well with FormX. You may have already noticed. The first one has a more titled angle while the other has higher background noise. Images submitted to FormX will go through several image enhancing processes before they are submitted to OCR services and later fed to AI models. With references provided from the Anchors, images are warped before our AI models start pulling information.
By calling our APIs, extraction results can be obtained using any app. Navigate to the "API" tab, then copy both the form ID and your access token by clicking the buttons on the top right corner.
Try calling the API with curl. Copy our curl example, replace the corresponding placeholders then press enter! With the correct payload given, you will promptly get the results. Simply translate this curl command to whatever language your app is built with, and you'll have got FormX easily integrated.
If you'd like to learn more about the APIs, we have a complete documentation in the "API" tab.
You've now successfully extracted information from a set of documents that share the same format! One more tutorial to go!
