About re Recognition
Leading the way in pattern recognition technology
About Us
re Recognition has been serving the OCR, ICR and pattern recognition needs of clients for over 30 years with KADMOS technology. re Recognition sought to develop a more accurate and faster technology to decipher patterns, notably machine printed text (OCR) and hand written characters (ICR) to meet the needs of a global client base.
Most OCR and ICR technologies use ‘Support Vector Machine’ (SVM) algorithms or ‘Neural Network’ (NN) based learning. KADMOS is different. It is based upon a proprietary algorithm that uses “discriminatory entropy”. This technically superior approach to both ICR and OCR outperforms both SVM and NN working alone or combined.
The proprietary feature extraction is based on the same mathematics as Computer Tomography and captures up to ten times more data points in the symbol or character that needs to be recognised. Then, sophisticated algorithms, embedded in the technology, determine which are the most significant. This means intelligent, accurate decisions can be made fast.
In development KADMOS technology has undergone supervised learning. Before it is allowed to make decisions it needs to see tens of thousands of samples of each character or glyph, then has to practise, practise and more practise; comparing each glyph with every other glyph in the world until it knows what are the obvious and the subtle differences between them. It is this intelligence that makes KADMOS technology so superior at recognising machine print and hand written characters.
Leading the way, the re Recognition team has been able to develop technology that does not need to rely on dictionaries, enabling fast processing speeds in over 200 languages. This passion and commitment to multilingual support means more complex non-Latin texts such as Cyrillic, Greek, Hebrew, Thai and Tamil can be supported. Additional languages are always a feature of the product road map.
Our customers choose re Recognition not only for the exceptional accuracy and speed it can offer them, but also because of our collaborative approach to complex product development. This hands on approach during set up and our highly regarded peer-peer support model means our customers have remained loyal to us.
Partnership
When needed, our approach is to work in partnership with our clients to support them in developing solutions for bespoke, complex ICR and OCR challenges.
Out team has significant experience of the ICR and OCR needs across a range of markets from finance, education, healthcare, document management and more. With access to our technical specialists, consultants and partners, we can support you in the quest for sophisticated ICR and OCR capabilities.
Get in touch
To learn more about re Recognition technology and to request a free trial
re Recognition Story
Childhood & GDR
At a very young age, I was taken from the Federal Republic of Germany to the GDR—without any possibility of resisting. My father was a doctor, my mother a doctor. It was therefore clear that in the workers’ and peasants’ state I had little chance of being allowed to study. My dream profession was naval engineer; I wanted to see the world.
Chance came to my aid: at the 3rd International Mathematics Olympiad, I achieved top placements in all competitions within the GDR. However, at the international round in Poland, my streak ended—there, knowledge beyond the school curriculum was required. And I had never studied mathematics outside of school.
Nevertheless, this opened up an opportunity: I was able to apply for a mathematics degree at the University of Jena. In my third year, I was the first to give seminar lectures on character recognition (early IBM devices, Prof. Steinbuch “Automat und Mensch”, …) and neural networks (IEEE PAMI, McCulloch & Pitts, …).
Studies & Zeiss
About two years after completing my studies, I moved from the university to the Zeiss works. The reason: university staff in Jena were not allocated housing. Shortly thereafter, I completed my dissertation. It was considered too valuable to fall into Western hands and was therefore locked away in the university library’s safe.
During that time, a former classmate from secondary school contacted me. He had studied textile machinery engineering in Karl-Marx-Stadt and was working on his dissertation. His topic: quality control for fiber fleeces. His approach: scanning using parallel lines at different angles and counting their intersections with the fibers.
He lacked some mathematical background, which I was able to provide. In doing so, I realized that this exact scanning method, combined with the mathematics behind it, was ideally suited for character recognition. I presented this approach to a high-level panel at the Zeiss works. The response: “Character recognition is being handled within the framework of the Council for Mutual Economic Assistance (Comecon) by the Belarusian Academy of Sciences in Minsk.” I did not want to move to Minsk. So the topic was shelved for the time being.
Escape & Imprisonment
Due to my extensive family ties in the West, I had no prospects for professional advancement in the GDR. When an opportunity arose to escape to the West, I took it immediately. Mistakes were made—by the escape helper and by myself—and the attempt failed.
Later, in my Stasi files, I read how my former boss—an informant—denounced a colleague who had remarked, “Pretty stupid to get caught.” The pre-trial detention in Gera is a story in itself. Just this much: I was likely being prepared as one of the exchange spies for the Stasi agent Guillaume, the exposed associate of Willy Brandt. The interrogation protocols, the entire process: lies, lies, lies.
After pre-trial detention came imprisonment in Brandenburg. The cells there were about three times as overcrowded as they had been during the Nazi era. Where Honecker had once been alone in a cell, there were now three inmates. Work consisted of winding electric motors in three shifts.
There was time to think—and to work on feature extraction for character recognition. Since any unapproved notes were regularly confiscated, I hid important data in bed frame pipes, cabinet locks, and similar places. At times, a dozen fellow inmates helped with calculations—including five convicted murderers. A belated and very sincere thank you to all of them.
Everything was done without calculators, using only pencil and paper. One fellow inmate, a mathematician, even developed an inverse formula to reconstruct the outline of a character from intersection shadows at different angles: (f + f”) / 2.
“The foundations were laid in a Stasi prison in the GDR.”
New Beginning in the West
In the spring of 1979, I was bought free by the Federal Republic of Germany—for 250,000 DM, as a Krupp manager later told me. Immediately afterward, I wrote down everything we had developed in prison—over three days, filling 30 pages. I was not allowed to bring anything with me from prison to the West.
My father, who had worked for some time as a doctor at the Buchinger Clinic in Überlingen on Lake Constance, had met Peter von Siemens there. He sent my 30 pages, along with an explanatory cover letter, to Peter von Siemens.
Siemens / CGK
To understand what followed, a parallel storyline must be considered. In the mid-1970s, Nixdorf decided to build mainframe computers. For this purpose, parts of AEG/Telefunken in Constance were acquired. The workforce was expanded from around 600 to 2,000 employees. Two years later, it became clear that the expanded team was incapable of producing competitive mainframes. The consequence: the Constance plant faced bankruptcy.
To salvage what was possible, the German federal government urged Peter von Siemens to acquire the plant for one Deutsche Mark. This led to the creation of the Computer Gesellschaft Konstanz (CGK) within Siemens.
CGK’s product portfolio included large machines for reading lottery tickets as well as handheld readers for machine fonts. After the separation from AEG, CGK had lost its scientific support. Peter von Siemens therefore forwarded my work to CGK.
During the job interview, I had the impression that none of the CGK managers understood my work on character recognition. Nevertheless, I was able to negotiate six months to further develop and test my approach.
Breakthrough
Another stroke of luck: at exactly that time, CGK was preparing the new machine-readable ID cards. Everything had to be completed within six months. The problem was not the machine-readable coding lines (OCR-B), but the processing of approximately 60 million application forms.
The result: my classifier was twice as good as the existing one—cutting the error rate in half.
Technology
To achieve the required recognition quality, several standard pattern recognition methods had to be modified or abandoned. One of these was the Bayes classifier. I demonstrated that significantly better results could be achieved using a set of base classifiers.
Crisis & Self-Employment
In 1982, another economic crisis occurred. Siemens decided to reduce its global workforce by 7%. While this was manageable elsewhere, the low turnover in Constance made layoffs necessary.
KADMOS & Success
In 1986, I developed new plans. A software-only solution for the IBM PC was created. The product name “KADMOS” was established, and the first customers from the United States placed orders.
International Expansion
After relocating to Switzerland, we were able to gain international customers such as SAP and the German Federal Printing Office. New products followed, including signature verification and an add-on for Tesseract, significantly improving speed and accuracy.