Timeline for Writing a word with straight lines
Current License: CC BY-SA 3.0
15 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 25, 2016 at 21:30 | comment | added | Martin Ender | @mbomb007 Possibly, but getting the spec tight enough seems tricky. I think I'll be staying away from popularity contests for the foreseeable future, and I'm not sure this would work particularly well with an objective image similarity scoring (also we already have a "draw this image with nothing but lines" challenge). | |
| Apr 25, 2016 at 21:28 | comment | added | mbomb007 | @MartinBüttner Perhaps this would make a good challenge on PPCG. | |
| Apr 24, 2016 at 21:29 | comment | added | Eric Towers | You might also consider choosing lines according to the kernel density of the Hough transform. I.e., compute this transform to get the density estimate of lines in the entire word and for each letter. Use the transform of each letter to stencil the word's transform (so that only lines in this letter are used in this letter, but their slopes tend to be similar across all the letters) and select lines from that transform. Then, with high probability, place 1 or 2 endpoints on the boundary of the letter. I'd spend time on this, but ... grading. | |
| Apr 23, 2016 at 15:03 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 918 characters in body
|
| Apr 23, 2016 at 14:53 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 2 characters in body
|
| Apr 23, 2016 at 14:51 | comment | added | Akiiino | @MartinBüttner Hmm... The problem seems to be with the lines parallel to edges; maybe, choose one point on border and another randomly inside the letter? I would try that myself if only I could understand the arcane magic you used; sadly, I'm nowhere near that level of understanding yet. | |
| Apr 23, 2016 at 14:41 | history | edited | Martin Ender | CC BY-SA 3.0 |
added 690 characters in body
|
| Apr 23, 2016 at 14:36 | comment | added | Martin Ender | @Akiiino Oh, that's a good idea, let me give that a go. | |
| Apr 23, 2016 at 14:28 | comment | added | Akiiino | The letters in the original picture are more pronounced due to the fact that their borders are much more visible: a lot of lines start/end exactly at the borders. Maybe choose only points on borders and extend only some percentage of lines (like 50%)? | |
| Apr 23, 2016 at 14:24 | history | edited | Martin Ender | CC BY-SA 3.0 |
deleted 37 characters in body
|
| Apr 23, 2016 at 14:21 | comment | added | Martin Ender | @J.M. Another idea might be to choose a fixed slope and put lines of that slope through individual random points. There will then be more lines along parts of the letter which are parallel to the slope than those which are perpendicular. Do that for a few different slopes (or vary them slightly about some mean slope) and you might get nice results as well. | |
| Apr 23, 2016 at 14:20 | comment | added | Martin Ender |
@yode Interesting, I tried applying DiscretizeGraphics directly to Text earlier but it didn't work without _Text pattern. That simplifies things, thank you.
|
|
| Apr 23, 2016 at 14:18 | comment | added | J. M.'s missing motivation♦ | "maybe prioritising those with angle close to ±90 degrees" - the route that requires least programming effort would be something like rejection sampling; that is, the segment picking can be biased towards vertical lines, and only allow less steep lines, say, 10 % of the time. | |
| Apr 23, 2016 at 14:18 | comment | added | yode |
Upvote.Or we can simplify your textRegion to DiscretizeGraphics[Text["MUSEUM"],_Text,MaxCellMeasure->0.1]?
|
|
| Apr 23, 2016 at 14:00 | history | answered | Martin Ender | CC BY-SA 3.0 |