The competition organiser, the Odessa National Academy of Food Technologies, invited more than 100 specialists from 20 countries around the world to join the jury. There were 239 projects to be evaluated in five competition categories including food science, economics, IT, robotics, automation, energy and ecology and environmental protection.
Students from Lodz University of Technology submitted a project entitled "FriML - Music generation using machine learning". It was placed in the most populated category "IT, automatics and robotics", in which 69 projects from 9 countries were submitted for a two-stage assessment.
The international jury composed of representatives from Poland, Angola, USA, Benin, Georgia, Azerbaijan, Belarus, Kazakhstan and Ukraine evaluated the project of three students - Ewa Kłapcińska, Wiktor Kania and Mateusz Groblewski from Lodz University of Technology - the highest and awarded it first place in the finals. The awarded application was created under the supervision of dr. inż Tomasz Jaworski and dr. inż Piotr Duch from the Institute of Applied Computer Science, supervisors of the Main student research group, in which students develop their skills and knowledge.
Briefly about the project
FriML is a music generation tool that uses machine learning algorithms and is available as a browser-based application. A clear interface allows the user to generate melodies, belonging to different music genres (pop, rock, classical), with different keys and for different instruments.
According to the students, their application has many potential applications. For example, it can be used by video game designers to add sound to their creations. The project is a solution for generating new melodies with just a few mouse clicks.
The solution is based on LSTM recurrent networks, used for tasks that require prediction of data sequences on the basis of a certain context. This includes text classification, speech and image recognition, sentence generation and music.
You can test the application at https://friml.netlify.app/. This is a demo version - the generated melodies last a few seconds and can only be saved in MIDI format.