Skip to main content

Timeline for Sandbox for Proposed Challenges

Current License: CC BY-SA 4.0

27 events
when toggle format what by license comment
Jun 12, 2022 at 18:33 history edited pxeger CC BY-SA 4.0
added 10 characters in body
Jun 17, 2020 at 9:03 history edited CommunityBot
Commonmark migration
Jul 4, 2018 at 4:55 history edited Jo King CC BY-SA 4.0
re-added sort-by notice
Jul 2, 2018 at 13:25 history edited r3mainer CC BY-SA 4.0
Moved to secret santa thread
Jul 2, 2018 at 12:06 comment added r3mainer @user202729 OK, done :-)
Jun 26, 2018 at 16:51 comment added r3mainer @user202729 No. What's that?
Jun 26, 2018 at 16:22 comment added user202729 Do you know about the Secret Santa's Box?
Jun 26, 2018 at 14:20 history edited r3mainer CC BY-SA 4.0
Update re image availability
Apr 25, 2018 at 15:06 comment added asmaier and: dropbox.com/s/buyg5swn0o0und2/SnakeEyesFinalPres.pptx?dl=0#
Apr 25, 2018 at 14:55 comment added asmaier See also: markfickett.com/stuff/artPage.php?id=389
Dec 26, 2017 at 16:19 history edited user202729 CC BY-SA 3.0
Add the notice because this is the most up vote and not deleted challenge.
Jun 14, 2017 at 17:13 comment added Santiago Benoit Seems like a perceptron could do the trick
Nov 3, 2016 at 18:48 comment added Martin Rosenau JPEG is an image format which is very hard to read when not using a special library. "Divide the size of the program by 100" is not a good idea when allowing programs having less than 50% correctness: A program simply writing random numbers will already have something about 10% correctness and its possible to do this within less than 1000 bytes (-> positive score); a program really evaluating images surely will need more than 10000 bytes (-> negative score).
Sep 23, 2016 at 16:42 comment added Magic Octopus Urn Ooo... a clustering algorithm question, this I like +1.
May 6, 2015 at 21:15 history wiki removed Martin EnderMod
May 3, 2015 at 15:37 history made wiki Post Made Community Wiki by Martin EnderMod
Jan 13, 2015 at 10:34 comment added r3mainer @trichoplax Yes, they would be standard right-handed dice. I can't see the point in announcing a subset of images for testing, since this would only benefit latecomers
Jan 12, 2015 at 23:37 comment added trichoplax is on Codidact now If the handedness/chirality of the dice is also guaranteed to be consistent then identifying 2 visible faces will uniquely determine the 3rd visible face, as a double check or to fill in unreliable data. I think it's worth stating explicitly whether the dice will be of consistent handedness. People can of course work this out for themselves by examining the image set, but it would be good to know for certain whether that is meant to be a reliable feature of the images.
Jan 12, 2015 at 23:35 comment added trichoplax is on Codidact now Can we assume these are all standard dice, so any two opposite faces add up to 7? This would allow double checks such as 3 and 4 can never be on adjacent faces, and 6 and 1 can never be on adjacent faces.
Jan 12, 2015 at 23:33 comment added trichoplax is on Codidact now You could post 1000 cases and then announce the subset of 100 that will be used for scoring.
Jan 9, 2015 at 22:10 comment added TheNumberOne @squeamishossifrage I meant so that people can generate more test cases if 100 isn't enough.
Jan 9, 2015 at 20:16 history edited r3mainer CC BY-SA 3.0
added 66 characters in body
Jan 9, 2015 at 20:15 comment added r3mainer @TheBestOne The Python script doesn't do anything exciting. It just places two dice in the camera's field of view with randomly selected faces pointing up, then gives then a random z-axis rotation from 0 to 2π (repeating the process if the two bounding boxes intersect). I added a little devilry at the post-processing stage (pincushion distortion, glare FX and depth of field blurring) to better simulate the output of a real camera and foil solutions that expect images with perfect geometry, but none of this takes place in the Python script.
Jan 9, 2015 at 20:02 comment added r3mainer @MartinBüttner Good points. I'll re-tag this as a code challenge. I'm trying to come up with a scoring system that prefers innovative hacks over long-winded but 100% correct solutions; if you can suggest any improvements then please do. I've just generated a set of sample images. In most of them, the two dice are non-overlapping, so it should be possible to score reasonably well without going overboard on the segmentation.
Jan 9, 2015 at 13:36 comment added TheNumberOne (b) Maybe you could post your script in addition.
Jan 9, 2015 at 11:20 comment added Martin Ender a) Either one or two dice seems fine to me. b) You should be able to upload the zip file into a Gist on GitHub. c) As far as I can, this is a code challenge and not a code golf. The golf score seems negligible in comparison to the accuracy score, and will probably serve mostly as a tie breaker. (And I think that's good.)
Jan 9, 2015 at 11:02 history answered r3mainer CC BY-SA 3.0