Hacker News Reader

This is pretty neat - can anyone give any insight/guesses into how this was done?

Well, the stupid way to do this involves having a stand with many cameras mounted in a globe shape, such that the viewing angles of all the cameras overlap. This is expensive. The more elegant solution is to use half-globe mirror (or two) to get 360 vision with 1 camera:

http://en.wikipedia.org/wiki/Omnidirectional_camera http://www.cis.upenn.edu/~kostas/omni.html

This is getting very popular in robotics applications because it's cheap and a simple way to get 360 surroundings data. From this, you can localize and / or detect obstacles.

Here's the mirror way of doing it:

http://www.360video.com

There's a link on the home page showing what the uncorrected video looks like (it's a giant video donut).

You get better image quality (e.g. no stitching errors), but lower resolution, because the whole environment has to fit in the same 1080P frame.

Maybe this picture will help you: http://www.creativeapplications.net/wp-content/uploads/2010/...

Now you have different options to display it. One is to map it on a 3D object and place a camera inside the object.

More info and movies: http://www.cornflex.org/?p=1

Reply on news.ycombinator.com