User verification

While developing a service for recognizing IDs and recognizing faces, we were surprised that our interlocutors had difficulties understanding the capabilities of the advertised products. They saw that Verigram recognizes a document in a video stream, compares images from an ID and selfies, but they did not understand what business process to integrate these technologies into. Even more, surprise arose when we realized that the interlocutor was an employee of a financial organization working with individuals.

We were perplexed. It was tempting for us to start explaining the benefits of automating the onboarding process. In the beginning, we did this, so we used the time ineffectively - we talked a lot and did not exchange contacts for subsequent communication.

Shiny terminal

The idea to make a biometric verification demonstration stand in the form of a clear terminal for everyone came after participating in several events. The potential client gave us 3-4 minutes of time to pitch, after which he politely moved on. Preparing and polishing the sales speech did not help much. Lack of interest in the eyes and unwillingness to exchange contacts was vivid evidence of this situation.

A visual example of our pitch for visitors was necessary to show. There was an idea to walk in defiant clothes with a presentation in hand and show pictures to attendants.

As a result, we stopped at a catchy stand. Such branded designs are found at the front of electronics stores in every mall. In 3 days, we threw in and approved the rack layout based on the size of the equipment that we planned to use. We wanted to get such a beauty with a compartment for reading and recognizing identity cards.

For several days we were looking for masters who would quickly and for little money take up the order. We were lucky that the manufacturer of outdoor advertising was among our friends.

The rack came out at 195 centimeters and weighed 60 kilograms. It was covered with shiny black paint and built-in diodes to attract the attention of conference and exhibition visitors from afar. It turned out like this:

What was inside?

There was simple equipment under the hood of the rack:

ASRock DeskMini 110

11.6-inch POS Monitor TVS LP-11R6

2 web-cameras Defender G-Lens 2577

The software was launching automatically and ran on a mini-PC. Services for text recognition and facial recognition were installed without problems.

After a while, we understand that the Nvidia Jetson Nano would perfectly cope with our artificial neural nets, further reducing the cost by 2.5 times. But the time for manufacturing was running out, and it was required to solve the issue of power supply and cooling of the device.

Service workflow from the user's perspective

The user places the identity card in a special compartment in the terminal for documents. The program automatically reads the text data from the face of the ID using a camera. In the case of correct recognition, the user is asked to look into the camera located above the screen to go through the verification process using facial recognition technology. The system automatically shoots after it detects a face by itself. The result of the comparison of the two faces and the recognized text of the identity document is displayed on the screen.

Result

Such a rush of potential customers was not expected. It was most pleasant to hear such conversations:

- I told you about them. Give me an ID. I'll show you how it works.

Pause.

- Did you see how quickly it worked?

What are the benefits?

To launch a biometric terminal or self-service kiosk, you need to solve the following tasks:

Determining the location for a biometric kiosk is the most difficult task. It is important to make sure that there is a client you need in the passing traffic and it is profitable to arrange an equipment installation point

Rent one square meter of space;

Connecting electrical power is the least difficult problem.

In which areas to apply document recognition and facial recognition?

There are many possible industries for these Machine Learning, Computer Vision, and Optical Character Recognition (OCR) technologies. After all, registration forms are filled out automatically, without human intervention.

Most popular verticals:

Telecom - sale of SIM cards

Finance - issuing money and issuing bank cards

Air and railway ticket offices - ticket sales, document checking when boarding flights

Hostels - guest registration and key collection

Insurance - sale of insurance policies

Border control - passing border control

Object recognition to improve process quality

Object recognition is a popular and undervalued niche