Music-Source-Separation

Abstract

In a world full of sounds, music is used as a connection for people around the world. This proposed project titled "Music-Source-Separation" is a deep neural network reference implementation for creating opportunities for researchers, audio engineers, and artists. The completion of the project will result in providing a genuine and clear concept of the sound and instruments. This project deposits music and permits users to separate pop music into four stems: vocals, drums, bass, and the remaining other instruments. The user can observe the sound from the sequence that is inserted from the system. A lengthy history of music separation has a scientific interest because of being thought of as an immensely difficult problem. For example, deep learning-based systems have been giving very meaningful separations which lead to increase interest commercially. MusicSource-Separation giving a reference implementation where a deep neural network is basically established. The project itself provides two main purposes. Firstly, accelerating all academic research in this field. Secondly, Improving the Bengali music community. The The Bengali music community has a bright history throughout the birth of the nation. Research for Bengali music has been very poor throughout the years. The artists also suffer from various problems during their careers and research has not been done for their work in their life. There is not a huge amount of collections from the past and also not enough resources created for future endeavors. Our work will try to affect the Bengali music community by creating a healthy and educative environment for all ages of people.