M6: multi-generator, multi-domain, multi-lingual and cultural, multi-genres, multi-instrument machine-generated music...

Area of Science:

• Computer Science
• Music Information Retrieval
• Artificial Intelligence

Background:

• Machine-generated music (MGM) offers creative applications but poses risks to human compositions.
• Effective detection of machine-generated music (MGMD) is vital for preserving the value of human-created art.
• Current research is limited by a lack of comprehensive datasets for MGMD.

Purpose of the Study:

• To introduce M6, a large-scale, diverse benchmark dataset for machine-generated music detection (MGMD).
• To provide a robust resource for advancing MGMD research and developing effective detection methods.

Main Methods:

• M6 dataset construction involved diverse generators, domains, languages, cultural contexts, genres, and instruments.
• Data was collected and analyzed, with baseline performance scores from binary classification models established.
• The dataset is provided in WAV format and is publicly available.

Main Results:

• The M6 dataset exhibits significant diversity, crucial for comprehensive MGMD research.
• Baseline models demonstrate the complexity of MGMD, indicating a need for advanced detection techniques.
• Initial analyses highlight the challenges and requirements for effective machine-generated music detection.

Conclusions:

• The M6 dataset is a valuable resource for the scientific community, facilitating progress in MGMD.
• Further research is needed to develop sophisticated methods for accurately detecting machine-generated music.
• M6 aims to foster collaboration and drive innovation in the field of music generation and detection.