Research on ML and RL applications in EEG analysis and brain-computer interfaces.
Research on autonomous driving and 1/10th scale race car, aimed at taking part in racing competitions.
More information on our current and past projects can be found on our blog.
Community support
As our contribution to Polish AI community, we put a lot of effort into supporting people who share our interest in machine learning. Here are some notable examples:
We have three times had the pleasure of being a partner of PL in ML: Polish view on Machine Learning conference dedicated to Polish achievements in the field of machine learning.
We organize "Machine Learning Beanfeasts" aimed at integrating Polish ML community.
We provided counsel to Mr. Mirosław Bartołd in his diploma thesis: "The application of an LSTM neural network model for identification of sign language words". The aim of the dissertation was the development of an algorithm capable of recognizing signs used in the Australian sign language, based on signal collected from a glove with mounted hand position sensors. Additionally, we published a blog post based on and developing the topics examined in the thesis.
We consulted master thesis entitled "Error-potential based reinforcement learning in brain-computer interfaces", prepared by Maciej Śliwowski, in the area of ML and RL. The main goal of the study was to create a model which based on the bioelectrical brain activity signal (EEG) detects error potentials (informing whether performed action was wrong) and uses it to increase the performance of the brain-computer interface. In the future, it may improve the quality of communication and interaction with disabled people.
We support Bartosz Topolski's master thesis titled "Image-Based Facial Expression Classification". The goal is to create an algorithm classifying basic emotions based on images of facial expressions. A secondary objective is to make predictions in real time. The work heavily utilises convolutional neural networks.
We provided computing power of our computer GUŚLARZ 9000 and supported Krzysztof Leszczyński in the development of network architecture. This thesis is entitled "Seizure prediction model reusable for new patients" and its goal is to create a machine learning model and test different methods of unifying EEG channels in terms of performance on new patients. Unifying methods applied in the thesis: 10-20 EEG system; selecting channels with the highest variance; autoencoder for unifying channels.
