European and Portuguese funding through projects allowed to initiate research, but complete applications (products) cannot be realised because of limited project durations. No other funding (e.g. University) means that other important lines of research cannot be initiated, and developed software cannot grow into products. Highly-trained students and postdocs are obliged to leave the lab and a big part of the knowledge and experience disappears. This implies that a state-of-the-art laboratory with one of the best computer parks and all investments in people and library etc is not fully employed. Sponsoring with the possibility of free licensing is a solution at little cost (see below) that can benefit both you and the lab.
Throughout the last years experience has been gathered in 2D and 3D imaging, including graphics and data visualization, with the possibility to parallelize algorithms on SMP systems (OpenMP) and small clusters (MPI). For the most demanding applications that require more than 2 CPUs, we're even developing libraries on top of MPI that hide all MPI stuff and that allow to parallelize data-parallel applications with a minimum of programming overheads. Possible products based on our experience are:
DIATOM IDENTIFICATION: a package for building image databases, extract contour and ornamentation features, plus computer-assisted (automatic) identification with different classifiers. CAN BE EXTENDED TO MICROFOSSILS ETC.
SURFACE RENDERING: a complete visualization pipeline for huge medical, acoustic and incomplete sonar data sets, consisting of interpolation, segmentation, labeling, triangulation, mesh optimisation and reduction, until VRML visualization or OpenGL-based interfaces with Gouraud shading.
REALTIME DIRECT VOLUME RENDERING: direct data visualization based on gradient detection, using colour and transparency lookup tables for alpha blending and Gouraud shading, plus a lot of other features, to be used on dual-CPU systems and Beowulfs (to be developed).
OPTICAL-SECTIONING MICROSCOPY FRONTEND: (to be developed) constrained inverse filtering of a stack of images taken at equidistant depth planes, plus surface detection for surface rendering. No surface detection for direct volume rendering.
TURNKEY CLUSTER DEVELOPMENT: we can contruct small Beowulfs with Linux and MPI plus optimized communication hard/software (MPI-over-SCSI and memcopy between dual processors).
MPI FOR THE UNINTERESTED: an optimized communication library (spmd_) on top of MPI that hides all MPI stuff for data-parallel applications. In the future the spmd_ calls to be inserted by the programmer can be inserted semi-automatically by a parser that recognizes a few directives similar to OpenMP.
SOFTWARE PARALLELIZATION: we can parallelize time-critical applications on dual+ SMP systems or on clusters running MPI (solutions for Silicon Graphics, Pentium and Athlon systems, the latter two with Linux of course).
VISUAL MODELS AND PERCEPTION: although nobody we know of is interested in this, we can do psychophysical experiments in controlled conditions for specific observation problems. We are also developing models for brightness and detection.
Portugal is still relatively cheap and there are excellent students interested in the topics mentioned above. Most of the infrastructure is already available. The costs involved are:
Personnel: a student costs 1250 Euros/month, a postdoc 1750. They do not pay taxes, which is one of the reasons for the much lower costs if compared to other countries.
Consumables: depends on the size of the project, something like 500 E/year.
Equipment: depends on the project requirements, but can also be a small contribution to maintain the computer park.
Travel costs: if you don't visit us, we'll visit you. Note: cheap charters to/from Faro international airport; modestly-priced hotel accomodations.
Overheads: for covering administration costs etc, normally 20% of all other costs.
NOTE: the administration will be done by CINTAL, a non-profit research organisation created by the University.
WHAT YOU GET: a free but not exclusive software license and source code that you can further tailor to your needs, if not already done by us. But if you have specific needs and want exclusive rights, well, contact us and we can talk about the situation. All is negotiable, although a research lab is not only a software house...
For all questions etc please contact Hans du Buf
Last update: April 2001, HdB.