The Audio Melody Extraction Annotation Initiative (AMEAI) is an initiative driven by the research community working in Music Information Retrieval and related fields. Its primary goal is to compile an open, large, representative and carefully annotated dataset for evaluating what is commonly referred to as Audio Melody Extraction algorithms.
In order to compile this dataset, the AME Annotation Initiative must first address the following topics:
- Define what should be considered as the "melody" in the context of AME evaluation
- Provide a clear definition of the Audio Melody Extraction task (i.e. what should an AME algorithm do)
- Describe the evaluation measures used to evaluate AME algorithms
- Determine the music material that should be included in the new dataset
- Define a clear protocol for generating new annotations
- Compile a list of supplementary resources related to AME
Each one of these topics has a dedicated page on this wiki. Once these topics have been addressed, the AME Annotation Initiative wiki can serve for coordinating and reporting on the progress of the annotation effort, and hopefully also as a general information source about Audio Melody Extraction.
Why do we need a new initiative?
This initiative stems from recent (2012) discussions on evaluation within the research community:
- J. Salamon and J. Urbano, "Current Challenges in the Evaluation of Predominant Melody Extraction Algorithms", in Proc. 13th International Society for Music Information Retrieval Conference (ISMIR 2012), Porto, Portugal, October 2012.
- The panel session on Evaluation Initiatives in MIR during ISMIR 2012
- Discussions on the MUSIC-IR mailing list.
How to contributeThe AME Annotation Initiative wiki has just been launched! (25/10/2012). The best way to contribute is by adding your ideas/proposals to the "discussion page" of each topic, or if you prefer, directly to the wiki page.
We already have several members sign-up to the wiki, I hope the number will grow with time and that the initiative will turn out to be a success. Let's see how this pans out!