Weronika Hryniewska
supervisor: Przemysław Biecek
When social distancing is necessary, artificial intelligence makes people's work easier. With the lack of human resources, every kind of help is important. The question is, how to ensure that a model is trustworthy and not biased? The answer is provided by explainable artificial intelligence techniques and careful verification of the process of preparing data and neural network for training.
The presentation shows how to determine if the model is biased and how to avoid it while training. The database used during the research consists of up-to-date images of the lungs of COVID-19 patients. Their limited availability makes them even more liable to bias.