22 Jan 2013


by robotconscience

An openFrameworks addon supporting the Flickr API. [view on Github]

Most people think of Flickr as primarily a social media tool, but I also like to think of it as an ever-expanding and virtually limitless source of data.  This addon looks like it still needs some work (compatibility with other operating systems and some additional features) but that’s what github is for, right?  I would be most interested in using this tool to pull down lots of images with similar keywords, especially landmarks (e.g. the Colosseum, Eiffel Tower, Cathedral of Learning).  From here, there’s surely a lot of cool image processing we could do to extract and combine information to form more complete depictions of these places.  If only we had some computer vision libraries…



by kylemcdonald

Alternative approach to interfacing with OpenCv from openFrameworks. [view on Github]

Oh look!  YES.  OpenCV is a great library for doing computer vision without having to reimplement it from scratch.  I would argue that it requires more of an understanding of computer vision than, say, faceOSC, to know what to use and when, but the upside is that it is much more open-ended in its possibilities and widely used in fields like robotics.  There are lots of other ofx addons that implement parts of openCV, but this one appears to do most of them.  I can imagine combining the SIFT / pose estimation features of openCV with the images acquired from Flickr to reconstruct common subjects in 3D, given enough images.  Technically, you really only need two to do one facade well, and every additional image just makes it better.  So what do we do when we have enough images to do a really good reconstruction?



by kalwalt

a simple wrapper for polyvox lib, to make voxel meshes with a Minecraft style or as Marching cubes [view on Github]

Make it with voxels!  This is much more fun than just reconstructing things to be photorealistic.  Adjustable voxel sizes make the resolution scaleable.  My dream would be to make my own ofxMinecraft plugin to take the output of these voxel reconstructions and import them into a Minecraft world.  In the meantime, this alone could provide instructions to minecraft architects to build the structures manually.  If we imagine combining these three addons, we could create a tool where a server full of Minecraft players could search for a landmark, see it rendered in voxels as reconstructed from crowdsourced Flickr images, and then view each layer individually in order to build it digitally from the ground up.

Like this!

(from Spyrobandi)