lakh-pianoroll-dataset - GitLite

KO | EN

gitlite — search

#javascript #hacktoberfest #react #ai #typescript #llm #go #machine-learning #rust #deep-learning #macos #awesome-list #vue #open-source

lakh-pianoroll-dataset

(★ 87)

A collection of 174,154 multi-track piano-rolls

파일 탐색기

최종 버전 다운로드 (.zip)

- FUNDING.yml
- labels.tar.gz
- - audio_player.html
  - icon_link.html
  - video_player.html
- - default.html
- - jekyll-theme-minimal.scss
- - musegan_logo.png
  - pianoroll-example-5tracks.png
  - pianoroll-example.png
- _config.yml
- comparisons.md
- dataset.md
- index.md
- labels.md
- representation.md
- statistics.md
- batch_synthesize.sh
- derive_labels.sh
- derive_lpd.sh
- download_amg.sh
- download_labels.sh
- download_lastfm.sh
- download_lmd.sh
- download_tagtraum.sh
- synthesize.sh
- - __init__.py
  - multitrack.py
  - plot.py
  - track.py
  - utilities.py
  - version.py
- binarizer.py
- cleanser.py
- collector.py
- config.py
- constants.py
- converter.py
- derive_id_lists_amg.py
- derive_id_lists_lastfm.py
- derive_id_lists_tagtraum.py
- derive_labels_amg.py
- matcher.py
- merger_17.py
- merger_5.py
- npz_to_audio.py
- README.md
- utils.py
.gitignore
LICENSE
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Source Code for Deriving Lakh Pianoroll Dataset (LPD)

> The derived dataset using the default settings is available
[here](https://salu133445.github.io/lakh-pianoroll-dataset/dataset).

1. Download Lakh MIDI Dataset (LMD) with the following script.

```sh
   ./scripts/download_lmd.sh
   ```

(Or, download it manually [here](http://colinraffel.com/projects/lmd/).)
2. Set the variables `LMD_ROOT` and `LPD_ROOT` in `run.sh` and variables in
   `config.py` to proper values.
3. Derive all subsets and versions of LPD, `matched_ids.txt` and
   `cleansed_ids.txt` with the following script.

```sh
   ./scripts/derive_lpd.sh
   ```

## Derive the labels for the LPD

> The derived labels can be found at `data/labels.tar.gz`.

1. Download the labels with the following script.

```sh
   ./scripts/download_labels.sh
   ```

2. Derive the labels with the following script.

```sh
   ./scripts/derive_labels.sh
   ```

## Synthesize audio files for the LPD

1. Install [GNU Parallel](https://www.gnu.org/software/parallel/) to run the
   synthesizer in parallel mode.
2. Synthesize audio files from multitrack pianorolls with the following script.

```sh
   ./scripts/batch_synthesize.sh ./data/lpd/lpd/lpd_cleansed/ \
     ./data/synthesized/lpd_cleansed 20
   ```

(The above command will synthesize all the multitrack pianorolls in
   the _LPD-cleansed_ subset with 20 parallel jobs.)