1/28/2024 0 Comments D flat major scale midiGiantMIDI-Piano includes 90% live performance MIDI files and 10% sequence input MIDI files. Each transcribed MIDI file contains the onset, offset, pitch, and velocity attributes of piano notes and pedals. Then, we transcribe those solo piano recordings into Musical Instrument Digital Interface (MIDI) files using a high-resolution piano transcription system. We apply a convolutional neural network to detect solo piano works. We further create a curated subset containing 7,236 works composed by 1,787 composers where the titles of downloaded audio recordings contain the surnames of composers. We search and download their corresponding audio recordings from the Internet. ![]() We extract the names of music works and the names of composers from the International Music Score Library Project (IMSLP). ![]() ![]() In this article, we describe the creation of the GiantMIDI-Piano (GP) dataset containing 38,700,838 transcribed notes and 10,855 unique solo piano works composed by 2,786 composers. However, there is a lack of large-scale symbolic datasets for classical piano music. Symbolic music datasets are important for music information retrieval and musical analysis.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |