rhythmic-pattern-transcription

Companion resources for the paper 'Transcribing Rhythmic Patterns of the Guitar Track in Polyphonic Music'

Downbeat processing

Install dependencies with uv:

uv sync

Run the downbeat processing script:

import numpy as np
from downbeat import process_downbeats

downbeats = np.sort(np.asarray([...], dtype=np.float64))
processed = process_downbeats(downbeats)

Dataset excerpts

excerpts folder contains audio excerpts from the held‑out test split used in the paper, 2 excerpts per difficulty category (simplified, intermediate, advanced, original). Each excerpt represents a 10-second clip of polyphonic music with guitar and has two associated files:

• {difficulty}_{i}_full_mix.ogg: The full mix audio file containing all instruments.
• {difficulty}_{i}_guitar_with_clicks: The isolated target guitar track before mixing with added click sounds at the strum onsets predictions. All predictions are made on other stem using the MERT-based model fine-tuned on other stems.