UA Vamp Plugins

The UAPlugins set is a library of VAMP plugins developed in the DRIMS project to perform onset detection and polyphonic transcription.
The methods used in this library were developed by Antonio Pertusa and José Manuel Iñesta, supported by the project DRIMS (code TIN2009-14247-C02), the Consolider Ingenio 2010 research programme (project MIPRCV, CSD2007-00018), and the PASCAL2 Network of Excellence, IST-2007-216886.

Download binaries (installation instructions)

(v2.0) 17/09/2015

• OSX binary
• Linux binary
• Windows binary

(v1.0) 23/01/2013

• OSX binary (32/64 bits)
• Linux binary (32 bits)
• Linux binary (64 bits)
• Windows binary (32 bits)

Download source code

• Source code repository (v2.0)

Compilation details can be found in the corresponding Makefile. The VAMP plugin SDK requires libsndfile to be compiled. Besides, the UAPlugins require boost and fftw3 libraries.

Onset detection

The onset detection algorithm relies on a one-semitone filterbank to detect abrupt energy changes in the harmonic contents of a signal. It was evaluated in MIREX 09 and described in the following paper:
Pertusa, A., Iñesta, J.M.
"Note Onset Detection Using One Semitone Filter-Bank For MIREX 2009",
MIREX 2009 - Music Information Retrieval Evaluation eXchange, MIREX Audio Onset Detection, Kobe, Japan (2009)

Multiple fundamental frequency estimation

The multiple fundamental frequency estimation algorithms are described in:
Pertusa, A., Iñesta, J.M.
"Efficient methods for joint estimation of multiple fundamental frequencies in music signals" EURASIP Journal on Advances in Signal Processing, vol. 2012, pp. 27 (2012)

The algorithm called EURASIP refers to the extended method of chapter 3 from this paper, which was evaluated in the MIREX 2008 contest with reference PI2. The method EURASIP+tracking corresponds to the same method but with tracking, as described in the same paper and already evaluated in MIREX 2008 as PI1.

For the Onset-based algorithm, combinations of those frames that are between two consecutive onsets are merged to yield the pitches within the inter-onset interval. This technique is close to segmentation, and it can obtain reliable results when the onsets are correctly estimated, as it happens with sharp attack sounds like piano.