Look into datasets - Githubissues

vvolhejn / thesis

ETH Zürich MSc Thesis: Accelerating Neural Audio Synthesis

Apache License 2.0

17 stars 1 forks source link

Look into datasets #11

Open vvolhejn opened 2 years ago

vvolhejn commented 2 years ago

DDSP:

violin: uses five performances from Musopen, namely John Garner's numbers II, III, IV, VI, VII
NSynth: 300k four-second samples of MIDI notes, but the notes are separated which is unrealistic

Tone Transfer (paper):

trumpet and saxophone: "home recordings"
flute: performances from Musopen. "Four pieces by Paolo Damoro (24 Etudes for Flute, Op. 15 - III. Allegro conbrio in G major, 24 Etudes for Flute, Op. 15 - VI. Moderato in B minor, 3 Fantaisies for Solo Flute, Op. 38 - Fantaisie no. 1, Sonata Appassionata, Op. 140))"

NEWT:

violin, trumpet and flute from the URMP dataset. Also used in Hierarchical Timbre Painting

RAVE:

strings: "an internal dataset composed of approximately 30 hours of raw recordings of strings in various configurations (monophonic solos and polyphonic group performances, with different styles and recording configurations)"
VCTK: "The Voice Conversion ToolKit (Yamagishi et al., 2019) is a speech dataset composed of approximately 44 hours of raw audio sampled at 48kHz, produced by 110 different speakers with various accents"

vvolhejn commented 2 years ago

Solos: A Dataset for Audio-Visual Music Analysis is used here: https://erlj.notion.site/Neural-Instrument-Cloning-from-very-few-samples-2cf41d8b630842ee8c7eb55036a1bfd6

vvolhejn commented 2 years ago

https://www.ismir.net/resources/datasets/ ISMIR's list of datasets

vvolhejn commented 2 years ago

From the ISMIR list, except the ones already mentioned above:

Good ones

VocalSet https://zenodo.org/record/1442513#.W7OaFBNKjx4

trained singers singing notes, scales and short melodies in different styles

GoodSounds https://www.upf.edu/web/mtg/good-sounds

recordings of notes and scales on classical instruments

GuitarSet https://guitarset.weebly.com/

guitars comping + soloing in different styles and keys

MARG-AMT http://marg.snu.ac.kr/automatic-music-transcription/

amateurs singing

Potentially useful

https://sigsep.github.io/datasets/musdb.html#musdb18-compressed-stems

not a single acoustic environment

https://gitlab.doc.gold.ac.uk/expressive-musical-gestures/dataset/-/tree/master/violin

one short melody snipped played by many players in different styles - maybe not diverse enough

http://mac.citi.sinica.edu.tw/GuitarTranscription/

single guitar notes with different techniques and amp settings