Евгений wrote > > 1) I have several ideas about how to implement this project. > Great. Demonstrating that you have a thorough understanding of the project idea by being able to expand on it with your own ideas is likely a benefit in the application process.
Евгений wrote > > The first thing that came to my mind is neural network. But this solution > requires a large training set of images. > An initial step in the project is likely to be to acquire a training set of road images with and without signs as all algorithms will need this at least to some degree. I believe there are a number of people in the community that have experimented with video based mapping and have collected videos while driving that can be used to create a training set. There are also a number of research papers on the topic and it might be possible to contact the authors to see if one can obtain their training and test videos if necessary. Евгений wrote > > Another way is to use OpenCV, which combines great number of various > computer vision algorithms. > As long as the license is compatible, which in the case of OpenCV with > BSD likely is, there is no direct reason not to use libraries where they > make sense and I would say is encouraged. I definitely had the use of > OpenCV in mind when proposing the project. The only problem with libraries > could be that it might make the task of setting up and using this project > harder for mappers than just downloading a josm plugin. Евгений wrote > > Last thing I think about is libraries like this > http://code.google.com/p/signfinder/, which developed for similar tasks. > Is it permitted to use third-party libraries in this project? > The signfinder code was a GSoC 2009 project. Where relevant it surely can be used. After all, the point of GSoC isn't a test or competition of how good a coder is, but to create something that is maximally useful to the mentoring organisation. Евгений wrote > > 2) In the application template there is a requirement to "provide a short > example of an OSM XML document that represents a polygon way". This task > is not related to the technologies used in the project of computer vision > based speed limit detector. Whether it necessary to provide OSM XML > anyway? And if not, do I need to provide any other test task to prove my > knowledge? > Imho, more importantly than the OSM XML is to have done a bunch of mapping in your home town to get a feeling for how this all works and where the current data collection and editing methods are tedious and therefore a video based speed limit / road sign detector would simplify editing work flow. After all, an important part of the project will be to integrate the computer vision based detector into an OSM editor to ensure it is actually useful to the mapping community. Never the less, writing the osm xml document should be a fairly trivial task and shows that you have some understanding of the internal data model used by the editors. Good luck in applying, Kai -- View this message in context: http://gis.19327.n5.nabble.com/GSoC-2012-Question-about-video-based-speed-limit-detector-tp5604815p5605047.html Sent from the Developer Discussion mailing list archive at Nabble.com. _______________________________________________ dev mailing list [email protected] http://lists.openstreetmap.org/listinfo/dev
