AI company Smart Engines produced 20% more scientific publications in 2022

Smart Engines
4 min readFeb 13, 2023

In 2022, Smart Engines employees presented 65 scientific publications in peer-reviewed journals and proceedings, with half of them in major international journals. The scientific results of Smart Engines have formed the basis of its own AI technologies such as high-speed document scanning SDK and new techniques for MRZ recognition.

In 2022, Smart Engines released 20% more scientific publications than the previous one. Of the 65 papers, 32 were published in scientific peer-reviewed journals, 20 were released in conference proceedings, while 13 were presented at major conferences in the field of AI, computer vision and image processing.

In June 2022, Smart Engines experts published an article in the Journal of Imaging with a data set (DLC-2021) for machine learning in the field of ID document analysis. It consists of 1424 video clips captured in a wide range of real-world conditions and focuses on tasks relating to ID document forensics. This paper will contribute to the development of neural networks able to detect signs of forgery or document presentation attacks.

In July 2022, Smart Engines artificial intelligence scientists collaborated in research which built a 3D model of the human olfactory bulb (OB), a small 0.5 cm section of the brain. The findings, published in the journal Tomography, will help research on neurodegenerative diseases and determine the causes of olfactory impairment of the COVID-19.

Photographic image of the human brain with two olfactory bulbs (bilateral reddish colored structures) and olfactory tract (violet-colored bilateral bundle of nerve fibers)

A study by Smart Engines researchers, published in Mathematics in November 2022, found considerable resonance. The article describes how neural networks sometimes find tumors that do not exist and thus mislead doctors using CT scans for diagnostic purposes. The company experts proposed a method to assess the reliability and accuracy of NN algorithms that subsequently allows them to identify the best one. Understanding and studying the underlying principles of neural network-based analysis of medical images is crucial for building strong and reliable AI-based technologies and solutions.

The arrow indicates the tumor, which must be detected by the doctor for a correct diagnosis. On the right image, due to noise, the tumor becomes poorly visible, and it will be difficult to make a correct diagnosis from this image.

“We are pleased that our scientists authored 41 papers last year and 65 this year. Ten of these were published in the most respected journals in the scientific community, namely those in the top quartile (Q1) of the Scopus database. Not many companies can boast such an achievement,” — commented Smart Engines CEO Vladimir Arlazarov, PhD in Computer Science.

The scientific results at Smart Engines formed the basis for its own software products, introduced in 2022:

• Full text document recognition

The Smart Engines team has established a new generation text recognition system. It is capable of finding a document in a photo or scan and recognizing all the text data in 102 languages. The speed of the system is also impressive. It recognises 15 pages per second on a server and completes processing of an A4 sheet in 3–4 seconds on a smartphone.

The technology allows you to read and analyze the text even if the document has already been folded. In other words, it transforms the smartphone into an instant scanner. This makes conventional flatbed scanners simply redundant.

English text recognition (OCR) with Smart Engines — YouTube

• Fully redesigned MRZ recognition

The Smart Engines scientists have fundamentally redesigned the AI-based algorithms and methods of the Machine Reading Zone (MRZ) detection. Recognition errors were reduced three times and the speed increased by 10%. MRZ scanning is used in case of state border crossing, flight check-in, accommodation booking.

• The Next Generation of ID Scanner

In 2022, the Smart Engines scientists presented the state-of-the-art identity scanner. The system implements a unique visual memory algorithm that has improved the quality and speed of automated document type detection. The system makes it possible to identify almost instantly the type of document among over 2,000 possible options.

Eco-friendly, fast, and secure software for European IDs | Smart ID Engine — YouTube

“It is our strong commitment to the “Science First” Principle that allows us to continually broaden and improve our list of solutions. We welcome the fact that business and science in our company coexist harmoniously and foster mutual development. In 2022, our global revenue increased by 43% and exceeded $5,8 million (€5,3 million), the number of recognitions sold increased by 91% and reached 216 millions”, — said Dr. Arlazarov.

References:
1. Meshkov, A.; Khafizov, A.; Buzmakov, A.; Bukreeva, I.; Junemann, O.; Fratini, M.; Cedola, A.; Chukalina, M.; Yamaev, A.; Gigli, G.; Wilde, F.; Longo, E.; Asadchikov, V.; Saveliev, S.; Nikolaev, D. Deep Learning-Based Segmentation of Post-Mortem Human’s Olfactory Bulb Structures in X-ray Phase-Contrast Tomography. Tomography 2022, 8, 1854–1868. https://doi.org/10.3390/tomography8040156

2. Smolin, A.; Yamaev, A.; Ingacheva, A.; Shevtsova, T.; Polevoy, D.; Chukalina, M.; Nikolaev, D.; Arlazarov, V. Reprojection-Based Numerical Measure of Robustness for CT Reconstruction Neural Network Algorithms. Mathematics 2022, 10, 4210. https://doi.org/10.3390/math10224210

3. K.B. Bulatov; E.V. Emelianova; D.V. Tropin; N.S. Skoryukina; Y.S. Chernyshova; A.V. Sheshkus; S.A. Usilin; Z. Ming; J.-C. Burie; M. M. Luqman; V.V. Arlazarov; MIDV-2020: a comprehensive benchmark dataset for identity document analysis. Computer Optics 2022, Vol. 46 (2), DOI: 10.18287/2412–6179-CO-1006

--

--

Smart Engines

A software development company with a broad expertise in computer vision, document processing and OCR, mobile recognition systems and AI. www.smartengines.com