Research in machine learning and deep neural networks requires time, money and extremely powerful software systems. FIT CTU provides not only students of Artificial Intelligence specialization with a computational GPU cloud based on the state-of-the-art NVIDIA DGX A100 software system worth over CZK 4 million. Thanks to this, students can, for example, simulate the airflow around an airplane wing or determine a person’s age based on facial scans.
Research in conjunction with information technology requires state-of-the-art equipment to make the results relevant and beneficial for practice. FIT CTU has therefore decided to support student AI projects with powerful NVIDIA DGX A100 software, on which the GPU cloud (Graphics Processing Unit) is based. Thanks to this technology, students can design and implement real AI applications that will find use in modern industries.
FIT students have already used the software for research in their final theses. Bc. David Horský, a graduate of the bachelor’s program, used the GPU cloud to simulate the airflow around the airfoil profile of an aircraft wing in his thesis – Flow modelling around airfoil with graph neural networks. Using artificial intelligence, David Horský accurately illustrated why the shape of the wing and how air flows over it is important for flying.
Ing. Martin Nykodem used the GPU cloud to solve the problem of automatic speech recognition in his thesis – Dictated numbers recognition model for an interactive voice response (IVR) company. Thanks to the software mentioned above, he created a machine-learning model to recognize numbers dictated during a phone call in the Czech language. The model significantly outperforms the current best solutions from Czech speech recognition companies, as well as solutions from Google and Microsoft. Martin Nykodem is currently working on founding a startup that will create an AI solution at the level of human interaction for companies to communicate with customers over the phone and train dialogues directly between the coordinator and the customer.
Bc. Filip Žďánský deals with the estimation of a person’s age using a three-dimensional facial scan in his thesis – Age prediction based on 3D facial scans. Filip Žďárský made the estimation using machine learning methods, especially artificial neural networks.
“I am glad our faculty can provide students with a high-end GPU computing infrastructure for research, which will ensure quality results for their research,” said doc. Pavel Kordík, Vice-Dean for Cooperation with Industry. “Thanks to the state-of-the-art technological background, our student projects result in real applications that have great potential for use in companies. Our goal for the future is to provide students with more of these systems,” he added.
FIT CTU also has a CloudFIT system based on the NVIDIA Tesla V100 platform, which students and researchers from the faculty can use. Thanks to the performance provided during the coronavirus pandemic, FIT CTU ranked 1st in the ranking of teams of Czech academic institutions that provided their computing power to the Folding@home project. The project aimed to help scientists find a solution to the COVID-19 disease. The faculty team provided the most computing power, worth nearly 100 million points per day, equivalent to the power of about 1,000 standard computers.
