Surface Reconstruction: Experimental Session
The ECG Workshop to be held in Sophia-Antipolis from the 10/21/02 to
the 10/25/02 will feature a session devoted to Surface
Reconstruction. this session will be split into two parts: one devoted
to talks ---the list of which will be broadcasted by Jean-Daniel
Boissonnat sometime soon, and one devoted to an experimental musing.
the details of the latter part go as follows.
Experimental session: generalities
the contenders will be given a CDRom containing selected models at the
beginning of the workshop. they will run their algorithm(s) during the
week in order to prepare the debriefing session to be held on the
24th. the debriefing session will consist of two parts:
- each participant will give a demo using the models proposed and will
emphasize the critical features of his/her algorithm(s).
- a debate will take place on the strengths and weaknesses of the
methods, the quality measures used to grade the reconstructions, the
challenges still to be gaped, the implementation issues, etc.
Experimental session: details
- any reconstruction algorithm is eligible: academic or industrial.
reconstruction methods not using the Delaunay-Voronoi machinery are
also welcome. (!) please send me email to register if you plan to
participate.
- participants are expected to come by with a laptop equipped with a
CDRom. each contender will be given a CDRom and will be free to run
the reconstructions / tune some parameters / fix some bugs / etc
during the workshop.
- the models' format will be an xyz file, that is a sequence of
coordinates as follows:
0 0 0
1 1 1
...
- the output of the algorithm should be a vrml or an off file.
- the algorithm should dump a logfile indicating some key facts:
- number of points discarded (if any)
- number of points created in the output (if any)
- number of connected components
- number of boundary edges
- number of non-manifold edges
- ...
- models will be classified into the following categories ---see
ECGWorkshop-ModelsClassification.ps.gz for a few examples:
- Academic examples
- Models with sharp features
- Surfaces with boundaries
- Models with severe topological issues (high genus, thin parts)
- Noisy models
- Huge point clouds (more than 1M points, to be reconstructed thoroughly or simplified)
- Sparse point clouds (less than 5k points)
- participants are encouraged to submit their own models. make sure, in
doing so, that the models you plan to send along are not protected by
any confidential agreement.
- two or three models will be chosen in each category. the choices will
make the difficulties as much independent from one another as
possible. (e.g. models with sharp features should not be noisy or
under-sampled.)
Frederic Cazals
Last modified: Thu Sep 26 17:07:55 MEST 2002 - Frederic.Cazals@sophia.inria.fr