Work, data and software page.

I am afraid there isn't much on this page yet. I am receiving an increasing number of emails about the data collection i used for my thesis experiements, i have decided to make most of the data i have used available online.

The first archive contains all the alphabet letters used in the various publications about object recognition in large databases of line patterns. Theses images can be used by anyone, as i created them myself, but i ask for a reference of my work in publications where they appear, are mentionned or have been used to compute the results. For both image database available for download below the following data files are available:

the original image in PBM format,
its canny segmentation,
its polygonised version (see papers or thesis for details),
its two dimentional geometric histogram representation,
its nearest neighboorhood graph representation,
its delaunay graph representation,
its constrained delaunay graph representation,
its two dimentional geometric histogram representation with nearest neighboorhood graph constraints,
its two dimentional geometric histogram representation with delaunay graph constraints,
its two dimentional geometric histogram representation with constrained delaunay graph constraints.

More information about the different file formats etc... are available in the ReadMe.pdf files provided within each archive.

Download the Alphabet Database zip archive (approx. 23MB).

The second archive is also public (from what i understood) and was provided to me by Myron Flickner IBM-Almaden Research Center. It is composed of 999 trademarks and logos.

Download the TradeMarks Image Database zip archive (approx. 93MB).

Should have any problem with the information or data on this page please contact me (don't forget to do the appropriate replacements!).