[CLAM] CLAM Success Stories

Luís Gustavo Martins lmartins at inescporto.pt
Mon May 17 03:46:54 PDT 2004

Hi Xavier and all!

This ideia seems a really nice opportunity to get to know the CLAM
community, and who knows, motivate interaction between people working in the
same topics or related areas (well... I assume that at least everybody using
CLAM is working with Audio! ;-))

So, here goes my two cents to your proposal:

I'm a PhD Student at FEUP (School of Engineering of University of Porto -
Portugal - www.fe.up.pt) working at INESC Porto (an R&D institute in Porto /
Portugal - www.inescporto.pt) in the area of  Semantic Analysis of Audio
Signals (you can visit my (still to be updated) webpage to get some more
detailed info about my work - http://telecom.inescn.pt/~lmartins).

> - Are you using CLAM for any kind of interesting project? Please comment
> it on the list even if it is only half baked.

Although I already heard about CLAM almost since its first public versions,
I've only started to get familiarized with CLAM some months ago (the
excelent docs and tutorials available at the CLAM webpage are mandatory if
you really want to get your hands into this framework).

My starting ideia with CLAM is to re-implement (and extend) in C++  my past
MSc project - a polyphonic Pitch to MIDI trancription system (the current
version is a MATLAB implementation; if you want, you can download it at my
webpage), and CLAM just seems to be the right framework for doing that
(specially because I use a very similar approach to SMS... the core of
CLAM's main processing classes).

Right now I'm migrating from MSVC6.0 to MSVC 7.1 (yes... I use MS Windows as
my base platform) and to the latest CLAM revision (0.6.1 - which has lots of
new improvements).

I expect to have some results using CLAM soon (no promises about this "soon"
statement,  though! ;-)).

> - Have you used other similar frameworks? Please report your experiences
> and relate them to CLAM.

I've been using Marsyas (a very nice C++ open-source audio analysis
framework developed by George Tzanetakis - http://marsyas.sourceforge.net/)
for a couple of years from now, for the development of some real-time audio
segmentation and classification tools.

Version 0.1 of marsyas has an approach that is somewhat diferent from CLAM.
While CLAM is more oriented to audio processing, marsyas 0.1 is more focused
on the analysis and classification of audio signals (or at least this is my
understanding of both frameworks).

Marsyas 0.1 provides some very usefull machine learning algorithms (Gaussian
Classifiers, GMM, K-NN, etc), feature extraction classes, and some nice
command line utilities (tools for creation and management of training sets,
cross-validation tools, etc) that seem to be "missing" in CLAM. On the other
hand, CLAM architecture and base classes seem to allow an easier
manipulation and processing of audio signals, and already provide support
for MIDI input/output, XML, multiplatform audio Input/Output, among lots of
other nice things not existent in marsyas0.1.

I've been in close contact with George Tzanetakis regarding Marsyas, and he
is now investing a lot deal of work in the new version of Marsyas:
marsyas0.2 (I pass the publicity! ;-)). This new version started more in the
line of CLAM (i.e. audio processing), presenting a completely new
architecture and implementing a lot of new functionalities, but in the most
recent revisions it is staring to inherit  some of the machine learning and
feature extraction classes of marsyas0.1.

Another important difference between marsyas(both 0.1 and 0.2) and CLAM is
that marsyas does not depends from external "third partiy" libraries, in the
way CLAM does, which of course, has advantages and disadvantages.

So, my options regarding which framework to use are: if I want to develop
some kind of audio analysis and classification application, I tend to base
my code around marsyas0.1. On the other hand, for a more signal processing
oriented application (such as my Pitch 2 MIDI app), CLAM appears as the most
logic choice. I really can't still say much about marsyas0.2 (I'm fighting
for some time to get my hands on that code!)... but maybe George can give
its input to this mailing list! :-)

Well, I hope this replies to some of your questions! ;-)



Luís Gustavo Martins
Porto - Portugal
lmartins at inescporto.pt

----- Original Message ----- 
From: "Xavier Amatriain" <xamat at iua.upf.es>
To: <clam at iua.upf.es>
Sent: Friday, May 14, 2004 1:44 PM
Subject: [CLAM] CLAM Success Stories

> Hi all,
> When browsing through the CLAM mailinglist archive you find out that
> most of the messages are about problems when installing and deploying
> The usual cycle seems to be:
> 1. User gets to know CLAM
> 2. User tries to install it on his/her system
> 3. User has problems and writes to the mailing list
> 4. User gets CLAM working (or not)
> 5. We know nothing else about the User
> This is of course a usual thing in similar situations. But we would like
> to make this list a bit more attractive and try in the future to build a
> sort of Clamming Community.
> Our first proposal is to invite users to write Success Stories:
> - Are you using CLAM for any kind of interesting project? Please comment
> it on the list even if it is only half baked.
> - Have you used other similar frameworks? Please report your experiences
> and relate them to CLAM.
> We know of some people using CLAM at McGill, Porto, CNMAT at Berkeley
> but we do not know what exactly (if anything) is going on. I think we
> could all benefit from sharing these experiences.
> PD. This is only the first of a few actions we would like to take in
> this line. We will keep you informed.
> xamat at clamdevel
> -- 
> /**********************************
> ********* Xavier Amatriain ********
> ****** Music Technology Group *****
> ** IUA- Universitat Pompeu Fabra **
> ****** www.iua.upf.es/~xamat ******
> ***********************************/
> _______________________________________________
> CLAM mailing list
> CLAM at iua.upf.es
> http://www.iua.upf.es/mtg/clam

More information about the clam-users mailing list