A Computational Analysis of Real-World DJ Mixes using Mix-To-Track Subsequence Alignment

This repository contains the code for "A Computational Analysis of Real-World DJ Mixes using Mix-To-Track Subsequence Alignment" Proceedings of the 21st International Society for Music Information Retrieval Conference (ISMIR), 2020. Taejun Kim, Minsuk Choi, Evan Sacks, Yi-Hsuan Yang, and Juhan Nam

This repository contains the code for:

1. Mix-to-track subsequence alignment
2. Mix segmentation
3. Subsequence DTW visualization
4. BPM change analysis

We are going to use the awesome mix below by Palms Trax since it is the author's recent favorite mix. data/meta/tracklist.csv contains the tracklist of the mix which is manually collected by the author from YouTube comments.

The figure below is the visualization of the subsequence mix-to-track alignment for Palms Trax's mix.

Installing python packages

NOTE: This repo is written and tested on Python 3.8.5.

You can install required Python packages using the code below:

pip install -r requirements.txt
pip install madmom==0.16.1  # madmom should be installed after installing cython

Downloading audio files from YouTube

python script/download.py

Running scripts

The scripts should be run in order below:

1. python script/feature_extraction.py extracts features and saves them under cache/ using disk-caching of joblib.
2. python script/alignment.py performs mix-to-track subsequence DTW and saves the alignment results.
3. python script/segmentation.py evaluates mix segmentation performances and saves the segmentation results.
4. python script/dtw_visualization.py (optional) saves DTW visualizations for all mixes under data/dtwviz/, but the repo already contains the visualizations.

Analysis notes

NOTE: The scripts above should be run before since the analysis notes require the results of the scripts.

1. note/mix_segmentation.ipynb analyzes the segmentation results.
2. note/bpm_change.ipynb analyzes the tempo adjustment.