[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Savannah-hackers] savannah.gnu.org: submission of PsiVision - Image Au
[Savannah-hackers] savannah.gnu.org: submission of PsiVision - Image Auto-Classification
Fri, 31 May 2002 17:51:40 -0400
A package was submitted to savannah.gnu.org.
This mail was sent to address@hidden, address@hidden
Tariq Rashid <address@hidden> described the package as follows:
Package: PsiVision - Image Auto-Classification
System name: psivision
This package does NOT want to apply for inclusion in the GNU project
The PsiVision project has the following aims:
* to produce an industrial strength, scalable, modular system for
automatic image classification, recognition, retrieval.
* to produce a system for quantifying the performance of this and
other similar systems.
* to produce excellent and extensive documentation to promote
the use of this project, to maximise its enjoyment,
to encourage a greater understanding of this and related
* to provide a friendly and fun environment for non-experts
(like myself) to learn about really cool facets of programming,
image understanding, databases, high availability computing,
mathematics, neuroscience, data mining, machine learning,
consciousness, philosophy, ... ;)
* proprietary with similar aims to psivision exist - they are
large, expensive and mysterious. we want to corrrect that.
* some free systems exist but are not entirely free -
eg academic use only, non-commercial use, etc
* some libre image processing projects exist - usually in the
form of libraries - but no complete solutions (eg no scheduling
of retrieval, no database systems, no integrated UIs).
* gnu gift is not entirely _automatic_ (it\'s interactive
query by example) and doesn\'t aim to be. psivision will encompass
all these methods eventually.
* the entire projects architecture will be developed independently
and evolves with comminity discussion and critical evaluation of
existing technologies such as IBM\'s CBIR.
* the entire project will be developed independently and not rely
on pre-existing components (eg GNU GIFT relies on viper project)
* project will consider the use of existing protocols if they are
totally open and free (possibly MRML as an example - see gnu
* code will be written using open and portable languages -
it is intended that the psivision backend will be in C.
the frontend (likely web based as well as cli) is likely to
be in python.
* all code is to be as portable as possible - follow example set by
OpenBSD\'s isakmpd project - with (eventual) compile targets for
net/free/open BSD and gnu/linux - sysdeps seperated off.
others may port to windows?
* initial testing and development on both gnu/linux and freebsd.
* it is difficult to predict what libraries might be required.
again, use only free libre code - if functionality is only
in a non-free library - we will write our own.
* likely libraries - graphics format loaders (libpng?),
maths libraries (libgsl?), possibly prolog/lisp libraries.
* algorithms either inspired from public sources or invented.
* as mentioned above, the documentation will comprehensively
cover design decisions and their reasoning. that includes
algorithms, sources and the use of libraries.
It does not exist yet but I\'m working on it.
Other Software Required:
difficult to predict - but will be entirely free and open libre libraries and
non-free functionality will be written from scratch.