Quantrium - ID-Card Data-Extraction Platform

The business problem

The verification of identity cards is an essential requirement for many routine applications such as employee onboarding, vendor empanelment, and customer registration. Traditionally, trained professionals inspect the documents to collect the data and validate the photograph. The time-consuming and error-prone process is inefficient, and scaling up the workforce to meet growing volumes is not cost-effective.

Processing speed and accuracy are critical operational goals. Automation with legacy OCR tools offers limited benefits, as they are incapable of handling the wide variation in the type and quality of documents.

Document designs and formats vary across regions and over time.

Extracting information from old, worn-out, or faded ID cards is a significant challenge.

Apart from text, photos and logos need to be recognized and validated.

Quantrium's solution

A fast-growing gig-economy company (with global footprints) was exploring technology-driven approaches to modernize the employee onboarding processes. The organization was recruiting thousands of executives every month to support business expansion and diversification plans. Manual scrutiny of documents was unequal to the task (and impractical).

Quantrium developed an AI-enabled solution to address the requirements. The platform facilitates the extraction of data, photos, and logos from a variety of ID documents. Built with state-of-the-art tools, the solution can accurately process documents of poor quality and readability. An API allows integration with downstream systems for instantaneous verification. An optional CAPTCHA validator module facilitates automated data acquisition from third-party databases. The solution effortlessly handles most common identity-proof documents such as the Driving License. The system can be trained on other documents to support future needs.

Handles soiled or faded cards and images with inadequate lighting

Accepts multi-lingual cards and documents of varied sizes and shapes

Add-on module to verify data against external, CAPTCHA-enabled databases

Architecture overview

Image Preprocessing

A suite of techniques, including PBT, gamma correction, and histogram equalization, enhances the quality of the image, making the text and graphic elements intelligible.

Information Extraction

An advanced text recognition engine extracts all the relevant details, including special characters. Images are extracted with Computer Vision-enabled techniques.

Data Verification

Advanced algorithms check the data consistency and similarity of photos across documents to facilitate a thorough assessment.

Integration with External Systems

APIs allow data exchange with external systems. Data can be fetched from CAPTCHA-enabled third-party databases with a custom-developed add-on module.

Implementation

Quantrium's design was based on powerful AI/ML tools and custom-developed algorithms for image processing and text recognition. The models were trained on thousands of documents, carefully selected to represent real-world applications.

Quantrium's cloud engineering team optimized the performance and reliability of the solution. Integration with the client's HRMS ensured seamless workflows.

Customer Experience

Quantrium's solution is used by the client for nationwide employee onboarding processes. Applicants' records are verified within seconds, enabling the company to meet time-bound staffing and training needs. The outstanding accuracy offered by the solution helps the company enhance customer trust.

Customer satisfaction
= 100%

Accuracy of processes
> 99%

Data extraction
< 60 seconds

Every day, we onboard scores of employees, and our HR team was cracking under pressure with tons of documents to examine. Thanks to Quantrium's solution, we now complete ID verification instantly. Kudos to Quantrium's team, and we look forward to engaging with them for our future needs.

- VP COO of the company