SynthSOD: Developing an Heterogeneous Dataset for Orchestra Music Source Separation

The SynthSOD dataset contains more than 47 hours of multitrack music obtained by synthesizing orchestra and ensemble pieces from the Symbolic Orchestral Database (SOD) using Spitfire BBC Symphony Orchestra Professional Library. To synthesize the MIDI files from the SOD, we needed to fix the original files into the General MIDI standard, select a subsect of files that fitted into our requirements (e.g.,  containing only instruments that we could synthesize), and develop a new system to generate musically-motivated random annotations about tempo, dynamic, and articulation. The code to replicate this process is available in our repository and all the details can be read in our paper. We have also published the code to train and evaluate the baseline and the pre-trained models in a GitHub repository.

We have also published the aligned score information for most of the pieces here.