About

Marsyas (Music Analysis, Retrieval and Synthesis for Audio Signals) is an open source software framework for audio processing with specific emphasis on Music Information Retrieval applications. It has been designed and written by George Tzanetakis (gtzan@cs.uvic.ca) with help from students and researchers from around the world. Marsyas has been used for a variety of projects in both academia and industry.

Featured Project

Analysis of Audio Features in Broadcast Sports Video

Multimedia Lab (Ghent University - IBBT) has been using Marsyas for the analysis of audio features in broadcast sports video. These audio features are used to detect semantically meaningful audio segments (e.g., cheering of the audience, commentary, whistles). This allows extracting specific events from the sports video that are useful for different types of applications (sports summarization and highlighting). - Chris Poppe, Multimedia Lab, Ghent University - IBBT, Belgium.

Featured Developer

Software Developer and Designer Thijs Koerselman

Affiliation : Ultrecht School of Arts

I'm a software developer and designer working with interactive media and sound. I hold an MA and BSc in Music Technology. After graduating in 2004 I got increasingly involved with programming. I have developed software for creative applications, live performance systems and art installations. Currently I work for the Utrecht School of Arts in the Netherlands, faculty of Art, Media and Technlogy, where we employ Marsyas in a project focusing on flexible and intelligent media repository software. Currently Marsyas is used for tasks such as music/speech classification and similarity matching. All content processing is done via a modular distributed pipeline framework, so additional algorithms can be easily plugged in. Other parts of the project include video analysis, data modeling and adaptive user interfaces. More info: http://im3i.in-two.com http://www.vauxlab.com

Featured Web Demo

CAL500 CAL500

The CAL500 dataset is a collection of songs curated by Doug Turnbull which has 500 songs of a variety of genres, each of which has been tagged with a variety of semantic tags by human listeners.

We recently used Marsyas to predict tags for each of the songs in this collection using a new technique called stacked generalization. Check it out at cal500.sness.net.

Featured Video

Somba

Samba rhythm creation using the Wii-Mote, Open Sound Control and Marsyas