cs229-audio-ml-project/musdb18-processed

Contents & fields

The dataset contains active stem segments extracted from MUSDB18, structured under extracted_stems/ with train/ and test/ subdirectories, each containing drums, bass, vocals, accompaniment, mixture folders, and a metadata/ folder with JSON files. Audio is 22.05 kHz mono. When loaded via Hugging Face datasets, each sample includes the following fields:

• audio——audio array and sample rate
• stem_type——type of stem (e.g., drums, bass)
• track_name——track name

Research uses

Suitable for music source separation, deep learning model training and evaluation, especially analysis of active vs. inactive audio segments.

This card was drafted from the source page; institution, coverage, time span, scale, fields and license are subject to the official page (pending human review).