This repository is an unmaintained, curated list of scientific articles, theses, and reports focused on applying deep learning techniques to music. It serves as a comprehensive resource for researchers and practitioners in the field of Deep Learning for Music Information Retrieval (DL4MIR), covering tasks beyond music generation, such as classification, transcription, and source separation.

How It Works

The project compiles a non-exhaustive list of academic papers, categorizing them by year and providing links to PDFs and, where available, associated code. It aims to cover a broad spectrum of DL applications in music, including but not limited to genre classification, onset detection, music generation, and source separation, differentiating itself from surveys that focus solely on music generation.

Quick Start & Requirements

• Installation: No direct installation is required as this is a curated list.
• Prerequisites: Access to academic databases or repositories for downloading papers.
• Resources: The README provides statistics on 167 referenced papers, 34 tasks, 55 datasets, 30 architectures, and 9 frameworks. Only 28% of the articles provide source code.

Highlighted Details

• Comprehensive coverage of DL applications in music, including Music Information Retrieval (MIR) tasks.
• Detailed tables summarizing articles by year, including links to PDFs and code.
• Statistics on research trends, tasks, datasets, architectures, and frameworks in DL for music.
• Links to numerous related resources, tutorials, datasets, and other "awesome" lists.

Maintenance & Community

The repository is marked as unmaintained, but contributions are welcomed. It was initiated and is primarily maintained by Yann Bayle, with contributions from several other researchers in the DL for music field.

Licensing & Compatibility

The project is licensed under the MIT license, allowing for free copying, modification, and distribution with attribution. It utilizes other projects like Pylint, Numpy, Matplotlib, and Bibtexparser, which have their own licenses.

Limitations & Caveats

The repository is explicitly stated as unmaintained, meaning the information may not reflect the absolute latest advancements. While efforts are made to link to published versions, some arXiv preprints might be included, requiring users to exercise caution regarding their peer-reviewed status. The list does not guarantee the quality or reproducibility of the cited articles.