Anselm, This is a fantastic idea!
I'd like to suggest that a BOINC client be developed so that the heavy processing can be done via volunteer cpu cycles. Micrsoft is already having to deal with the energy consumption issues of the CPU power needed to wide application of photosynth: https://mail.google.com/mail/?source=navclient-ff#inbox/11adaf6fc25718df A grid, volunteer model would allow for greater breadth of voxel modeling than I think even Microsoft can acheive with dedicated CPUs. Maybe we could encourage sites like Flickr to add Open Voxel Space specific tags! -Eric On Mon, Jun 30, 2008 at 1:32 PM, Anselm Hook <[EMAIL PROTECTED]> wrote: > I wanted to just share the observation that it might be time for us all as > a community to look at building a kind of open photosynth - what could be > called an 'open voxel space' map of the planet. > > Microsoft's photosynth project stitches together a series of arbitrary user > photographs of a scene, taken from different angles and perspective and > melds into a single seamless 3d image of that setting. The way this works > is that each image has notable feature points on it, and between any two > images there may be zero or more shared feature points. If you have enough > similar feature points then you can effectively say that these two images > overlap each other in some way. Given two photographs of a building, say > from two people standing at two different places, you can start to > re-constitute a 3d voxel model of that building from just those two > photographs. > > The goal would be to start collecting all photographs and building an open > 3d model of the photographed planetary surface of earth. Basically one > would be building a kind of open voxel space - a 3d model of our cities and > spaces - and this could help with other projects. > > The algorithms are not hard to use, there are open source implementations > (google SIFT) - and even if they algorithms suck right now they will improve > over time. It's mostly just a scaling problem; how and where to aggregate > or index or store the images. In fact I strongly recommend at least playing > with one of the open source SIFT implementations - it is a lot of fun and > gives you a taste for the possibilities. > > Generally I feel that the best data is more data. An Open Voxel Space > could help with lots of other problems. We define and attach labels to > streets as a way of doing a kind of manual position sensing. We of course > care to know about streets and paths because we cannot walk through walls. > > Good voxel data could help fix up bad GPS data among other things... and it > would help increase the usability of GPS data therefore and reduce the > necessity for thinking about streets and labels. With enough good data a > router would simply pick from the most common recent existing full gps track > between the two points... > > Also (and this is less firm, more speculative, but still seems marginally > relevant): maybe a voxel map of space; that wasn't just focusing on labelled > streets or paths, but on space in general, might also help with the puzzle > of better delivering real time and volatile data to people - "just in time > knowlege" - the "help I lost a kitten" kinds of stuff. It's unclear to me > why there isn't really a kind of real-time bartering service yet - perhaps > twitter is closest - but one that focuses more on discrete signalling for > very specific services; rather than just "saying"... Maybe a better map of > space could help - maybe the issue is that "distance" between things is > blocked by buildings in a way that street-maps don't quite convey and when > that is not clearly factored in it acts as a barrier to surmounting > distance? > > > > > _______________________________________________ > Geowanking mailing list > [email protected] > http://lists.burri.to/mailman/listinfo/geowanking > -- -=--=---=----=----=---=--=-=--=---=----=---=--=-=- Eric B. Wolf 720-209-6818 PhD Student CU-Boulder - Geography
_______________________________________________ Geowanking mailing list [email protected] http://lists.burri.to/mailman/listinfo/geowanking
