News for hackers. Stuff that matters.
Re:Image Triangulation Experiment by j2kun on Monday October 27, 2014 @05:19PM (#8518242)
I once tried implementing the classical Delaunay triangulation algorithm from scratch and got close to a good solution, but never could iron out some major bugs. I think of it as a testament to the difficulty of algorithms and computational geometry.
Re:Image Triangulation Experiment by minopret on Tuesday October 28, 2014 @05:49AM (#8520139)
There's a very clear exposition and implementation in the book and code library _The Stanford Graphbase_. It's reasonably simple for the computer. It's not so simple for the algorithm designer.
Re:Image Triangulation Experiment by HardyLeung on Monday October 27, 2014 @07:27PM (#8518749)
Neat. I have a related project called Polygonian -- http://www.polygonian.com -- runs on browser and requires Chrome Native Client that does something similar. Here's the output of a Lincoln portrait using similar point count:

http://imgur.com/lpKVILV

Re:Image Triangulation Experiment by spydum on Monday October 27, 2014 @09:15PM (#8519055)
seems like it's using gradients? for some reason looks less.. triagular that way
Re:Image Triangulation Experiment by HardyLeung on Monday October 27, 2014 @09:29PM (#8519094)
Yes, linear gradient within each triangle. It has a non-gradient mode but not turned on. The difference is minor though: Gradient = more "silky" and less triangular. No gradient = a bit more flat.
Re:Image Triangulation Experiment by wgx on Monday October 27, 2014 @11:14PM (#8519371)
This one's older, but a lot more fun IMO http://alteredqualia.com/visualization/evolve/
Re:Image Triangulation Experiment by bane on Monday October 27, 2014 @09:53PM (#8519162)
Ack, I thought this was a way to use two photos of a location from different angles to figure out where in the world it is.
Re:Image Triangulation Experiment by pimlottc on Monday October 27, 2014 @08:33PM (#8518935)
> "drop an image in the browser to triangulate it."

Am I the only one who finds this annoying? I would much rather be able to use any image on the web by submitting a URL than searching for something on my local drive.

Re:Image Triangulation Experiment by pornel on Monday October 27, 2014 @09:10PM (#8519043)
Unfortunately the platform makes the annoying solution easy, and the one you prefer hard.

D'n'd and forms give full access to the image data relatively easily, and it works completely client-side without server's involvement.

OTOH due to same-origin policy getting data from a random URL on the Internet is (intentionally) hard. "Simple" proxying via own server is tricky to do without accidentally building tool that could be abused for DoS or anonymous download of something that you wouldn't want to be associated with.

Re:Image Triangulation Experiment by pimlottc on Tuesday October 28, 2014 @01:26AM (#8519664)
Good points, I hadn't thought about the app being completely client-side and the restrictions that entails.

Still, in ye olde times, it was pretty common to accept image URLs. You might be able to do some fairly simple rate limiting on the proxy to help prevent abuse. Maybe the net was just a more innocent place back then.

Anyhow, if it only works with images dragged from the user's local computer, it should really say so. It's especially confusing on Chrome, where UI cues like the mouse cursor changing suggest that dnd between windows should actually work. It generates a drop event and everything, although it appears that the DataTransfer item just contains the image URL and not its actual contents.