The following article shows how to process ID documents automatically with myBiros. Specifically, how to automate theextraction of personalinformation. This service can be useful in many application scenarios, such as during customer onboarding remotely or in physical presence: in such cases, processing IDs automatically can streamline the customer registration process.‍

‍

Use case description

Identity documents (ID card, passport, driver's license, health card) fall into the category of structured documents. Therefore, it is possible to process ID documents with traditional rule-based and template-based solutions applied to the output of an OCR engine.

For the type of documents in question, however, there are several complications that do not allow traditional solutions to fully automate the use case. Such complications arise from:

• Document capture is not always guided. This can result in documents that are rotated and of low quality and therefore difficult to read. Such issues do not allow rule-based and template-based approaches to solve the use case in question;
• although the individual ID is easy to interpret, every single nation has one or more IDs. Therefore, reasoning about a global use case, we are talking about hundreds of documents that are different from each other. This means that by adopting traditional solutions, rules and templates are needed for each individual document in question.
• documents over time change format, this results in additional rules and templates to cope with the change.

The problems mentioned so far can be solved with innovative solutions that leverage deep learning techniques, thus obviating the creation of rules and templates. Let's see how myBiros solves this problem.

For simplicity, the article will discuss the use case related to the Italian ID card.

‍

Making a use case with myBiros

Thanks to the deep learning techniques employed by myBiros' technology, it is possible to zero in on the need for rules and templates, relying on a completely data-driven approach. The myBiros technology not only analyzes field position for extraction purposes but also leveragessemantic analysis,analysis of the document'sgeometry and finally its layout.

Building a use case with myBiros is very simple and consists of the following steps:

• Document collection;
• Data annotation;
• Training;
• Service release and performance testing.

‍

Collection of documents

The first step involves collecting a small sample of reference documents: 10 Italian ID cards. The collection of documents is necessary because AI algorithms for learning need training data to instruct the algorithm.

‍

‍

Data annotation

The annotation phase allows documents to be transformed into data that can be understood by the AI. Thanks to theintuitive no-code interface of myBiros, it is possible to specify the information of interest by simply clicking on the data to be extracted. In addition, myBiros' AI tries to suggest the information of interest to be extracted, speeding up the process.

‍

‍

Training

In this step, the algorithm learns from the information prepared in the annotation step. This step is fully automatic and in a few hours you will get the new trained model. Distinctive feature of the myBiros platform is that for the training phase you can choose: instruct the algorithm from scratch or choose one of the models pre-trained by the myBiros team on other domains to speed up the training and increase accuracy.

During training, model evaluation metrics can be monitored regarding the accuracy of the extracted data.

‍

‍

Release of services and performance testing

After the training is completed, the service can be tested through a simple interface that allows the results to be visualized. In this way, the performance of the model can be appreciated. The interface releases the new API related to the realized use case, showing some examples of possible integration.

‍

‍

Results and benefits

The result is an AI model capable of understanding the information of interest specified in training. This is all encapsulated in an API that can be easily invoked remotely.

The benefits of implementing such a use case:

• Decrease in errors due to manual reporting of information;
• Elimination of repetitive and alienating activities related to data entry;
• time and cost savings;
• more accurate and secure data;
• Speed of processes, made more streamlined and immediate.

Want to find out more about our solutions? Contact us and try our demo, we are waiting for you!

‍