In October 2017 four representatives of the medical.ml student research group participated in the HackYeah hackathon, the largest stationary hackathon in Europe. The group works under  scientific supervision of dr hab. Urszula Markowska-Kaczmar, prof. Pwr.. Our group was represented by a team: mgr inż. Filip Bachura, mgr inż. Gabriela Czachor, Piotr Janyst and mgr inż. Tomasz Wasyłyk. The event took place in Krakow. More than 2,000 people came to the event to choose one of the available competition challenges. However, the biggest challenge was time – the solution should be presented within 24 hours.

The tasks concerned various areas, for example: finances, transport safety or natural language processing. The aim of the participants was to develop a solution to one of the problems that have been reported by companies or state organizations that were the organizers of the event. These include: detection of financial fraud (Ministry of Finance), prevention of natural disasters, accidents during extreme sports (Lotto) or ensuring road safety (Luxoft).

Our team has taken up the challenge of developing an application designed to make transportation and driving easier and safer. The project used methods of deep learning and artificial intelligence. The team created two modules based supporting the process of driving a car: Steering Assistant and Driver Visual Aid. The role of the first one was predicting the suggested car’s steering angle based on the image in order to avoid obstacles. The second module also used the image to mark other vehicles and objects in the driver’s field of vision. They operated independently, in parallel, but the recognition was made on  the same image. Due to the lack of time to repeat the learning process during the competition, transfer learning was applied.

 

The proposed solution was awarded with the first prize in the challenge category  ‘safety’  and  outrivals 30 other teams.