Automatic Information Extraction from National Identity Cards using Deep Learning YOLO

Project Title

Automatic Information Extraction from National Identity Cards using Deep Learning YOLO

Project Area of Specialization Artificial IntelligenceProject Summary

Organizations need to handle their customers’ credentials by collecting personal information. National Identity Cards (NIC), Juvenile Cards (JV), COVID-19 Vaccine Cards, and other documents similar like that fall under this category. Manu-
ally handling such records is time-consuming and costly. I attempted to fix this issue by utilizing AI to automate this procedure. As a result, the built application will take images as input and output information about the consumer in an organized document that a computer can interpret. There are several open-source OCR wrappers available online that we cannot utilize directly for information extraction. It will be inappropriate to extract the desired information using direct OCR on such data. I am going to tackle the problem of manual handling of data by automatic working. Using the same procedure, we can use such automation in many other applications, similar that. Such new technology in real-world applications to save time and money is a good idea. 

Project Objectives

Project Description:

Technology is the scientific understanding that allows humans to develop, and invent innovations and automated systems. The progress in technology has made living simpler. Science and technology are inextricably linked; advancement in science leads to advancement in technology.  Advancements in technology lead to technical products and devices that benefit humanity. One such advancement which greatly benefits humans is automation, automation reduces time and errors. Automation in information extraction and organization is of the utmost importance in today’s digital world. The tedious task performed by a human to receive, check and extract information from any identification card is time-consuming and prone to human errors. Additionally, the process of arranging the extracted information requires more time and effort. The solution to this is to design a reliable system using AI to automate the whole process. The developed approach will take images as input and provide information about customers in an organized document for ease of understanding and reduction of time. To cater to my project for automation in Pakistan, I have focused my efforts on Computer National Identity Cards (CNIC). The information in CNIC is displayed in English and Urdu languages. So in my project, I am working on the identification and extraction of information in both English and Urdu language. I have successfully identified English and Urdu language texts in CNIC as well as English text recognition.

Project Implementation Method

OCR is an acronym for Optical Character Recognition. It reads text from images such as scanned documents or a photo. This technique almost turns any image with written text (typed, handwritten, or printed) into machine-readable text data. Here, I am going to design a Custom OCR that only extracts the information that I want to extract from the given image. There are two major building blocks of OCR;
• The first one is Text Detection, and
• The second one is Text Recognition

First and foremost, there is a need to extract specific text from images/documents. Often, depending on the situation, you don’t want to read the complete document, but rather only a portion of it, such as a credit card number, College Cards, name, amount, date from bills, and so on. It is difficult to detect the needed information, but thanks to deep learning, we will be able to selectively read text from a picture. Text detection, or object detection in general, has been a subject of intense study, driven by deep learning. Object detection, and in our instance, text detection,
may now be accomplished using one of two methods.
• Region-Based Detectors
• Single Shot Detectors
The primary goal of Region-Based techniques is to discover all the areas that contain the items and then feed those regions to a classifier, which returns the locations of the needed objects. As a result, it is a two-step procedure. It first
determines the bounding box and then the class of that bounding box. This method is thought to be more precise, although it is slower than the Single Shot method. Faster R-CNN and R-FCN use this approach. However, single-shot detectors predict both the bounding box and the class at the
same moment. It is significantly faster because it is a one-step operation. The Single Shot detectors do not work correctly when detecting small items. The SSD and YOLO detectors are Single Shot detectors. Basically, Text Detection is done using Deep Learning Model - YOLO. While English Text Recognition is done using Tesseract OCR and Urdu Text Recognition is done using another deep learning model.

Benefits of the Project

Throughout the years, technology has revolutionized our environment and our daily lives. Furthermore, aging technology has brought great tools and resources, bringing critical knowledge to our fingertips. Modern technology has enabled multi-functional gadgets such as the wristwatch and smartphone. Computers are getting more powerful, portable, and portable than ever before. As a result of all of these
changes, technology has made our lives easier, faster, better, and more enjoyable. As I used automation in designing a custom OCR API for automatic information extraction from Pakistani Identity cards. Such type of work was not available on Pakistani Identity Cards, so I proposed this system design. By using automation, we can design different systems for humanity’s ease. We can build different systems
for automatic MDCAT/ECAT Test checkers, custom OCR for passports, COVID cards, etc. Because we can use such new technology without any harm for making life easier, we should work on automation.

Technical Details of Final Deliverable

An API is designed that will do this work of automation. I am designing an API so that different organizations can use this application according to their requirements. Like some can design a mobile app, some can desktop applications, web apps, etc. 

Final Deliverable of the Project Software SystemCore Industry ITOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources