AILab@TUL
The Artificial Intelligence Laboratory at Technical University of Liberec
Who we are
The Artificial Intelligence Laboratory at Technical University of Liberec (AILab@TUL) has a strong experince with research and application of machine learning methods for processing of speech, image, and text data. In speech, this is mainly the task of automatic transcription, speaker diarization, or language identification from acoustic signals. In the context of image processing, the lab is primarily specialized in methods for object or person detection. In text processing, the researchers exploit extensive pre-trained language models, for example, in the task of automatic punctuation restoration or retrieval augmented generation (RAG).
Recent cooperations
These are the most recent cooperations and projects of AILab@TUL.
library system
We cooperate with Tritius company within the LibriBot project (MPO TWIST) to develop software services allowing for content-basd book recomendation and searching using natural language.
prediction of behavior for machine components
To goal of our project with LENAM company is to utilize modern machine learning methods to create models that can quickly and accurately predict the behavior of machine parts under various conditions.
SEARCHING in JUDICIAL DECISIONS
We have developed a software module enabling efficient, accurate, and semantically rich searches in large corpora of judicial decisions using vectorization models. The module is commercialized by the spin-off company nanotrix.ai.
Speech applications
These are the speech applications powered by technologies developed in coopertion with AILab@TUL.
Beey platform
Online tool for fast and accurate speech recognition and voice-to-text conversion at affordable prices. It allows the creation of high-quality subtitles with translation in a fine-tuned editor in just a few minutes
newton dictate
Professional voice dictation software, currently available for Czech, Slovak, Polish, Croatian, and Serbian.
my voice 2.0 + mydictate
Voice control of Microsoft Windows system and dictation tool for users with specific needs. The programme is recognised as a compensatory aid and is distributed by SILOU HLASU, z.s.
Commercial partners
nanotrix.ai
Our spin-off company helps us commercialize research results, e.g., the software module for searching judicial decisions using natural language. The core team of nanotrix.ai consists of experts from AILab@TUL, with many years of experience in research in speech, image, and text processing.
TRITIUS SOLUTIONS
Tritius is a software company based in Brno with over 30 years of experience in developing library systems. We cooperate with this company in the LibriBot project (MPO TWIST) to develop software services allowing for book recomendation and searching using natural language.
LENAM
Since its inception, LENAM company has provided top-quality services, particularly in the area of development support, mainly in the automotive industry . We help the people from LENAM to utilize modern machine learning methods to create models that can quickly and accurately predict the behavior of machine parts under various conditions.
JABLOTRON
The company is a global technology company with in-house development and production. Its history dates back to 1990. It specializes in home security and control systems, making life easier for millions of people worldwide. AILab developed for Jablotron an algorithm for motion detection with pet immunity.
RIETER
The company constitutes the world’s leading supplier of systems for manufacturing yarn from staple fibers in spinning mills. Based in Winterthur (Switzerland), Rieter develops and manufactures machinery, systems, and components to convert natural and manufactured fibers into yarns. AILab developed the Trash Level Monitor for Rieter, a neural network-endowed optical sensor for carding machines. It analyzes the presence of foreign particles in raw material and allows the achievement of the targeted fiber quality level.
NEWTON technologies
In various TACR projects, we have developed several ASR modules for the main European languges that are utilized in the Beey platform. Moreover, the voice dictation system NEWTON Dictate utilizes the recognition engine developed at our laboratory.
Academic partners
UNIVERSITY OF GRANADA
In the framework of cooperation with the University of Granada, the exchange of doctoral students takes place in particular. A number of joint publications have also been produced in this way.
NORWEGIAN UNIVERSITY OF SCIENCE AND TECHNOLOGY (NTNU)
We collaborated with colleagues from NTNU on the NORDTRANS research project, which was focused on the development of a technology for automatic speech transcription in selected Nordic languages.