« first day (927 days earlier)      last day (3558 days later) » 

1:05 AM
@blochwave. That's a "dog bites man" event; what I experienced was a "man bites dog" event :-)
 
 
1 hour later…
2:20 AM
After defining "a" as a simple utility function to convert JSON to associations, I'm quite pleased with the readability improvements allowed in v10 to the code to import category members from Wikipedia.
a[list_] := Replace[list, a : {__Rule} -> <|a|>, {0, \[Infinity]}]
categoryMembers[title_] :=
 Module[{url =
    "http://en.wikipedia.org/w/api.php?action=query&list=\
categorymembers&format=json&cmlimit=500&cmtitle=Category:" <>
     URLEncode@title},
  Join @@ NestWhileList[
     a@Import[
        url <> "&cmcontinue=" <> #[["query-continue",
           "categorymembers", "cmcontinue"]], "JSON"] &,
     a@Import[url, "JSON"], KeyExistsQ@"query-continue"][[All,
     "query", "categorymembers", All, "title"]]]
categoryMembers@"National anthems"
 
2:50 AM
Oh! Pressing tab for auto-complete and then pressing the down arrow will let you insert an argument template now. Very nice.
 
3:01 AM
@halirutan Yep, just ran into a pure function I needed to surround with parentheses.
 
3:20 AM
Ok, I updated the random premise seed selector to use 10,000 important Wikipedia articles instead of a dictionary. The results are definitely easier for your mind to latch onto, but still take thought to find a fun way to stitch them together. 3 articles and 2 tropes.
a[list_] := Replace[list, a : {__Rule} -> <|a|>, {0, \[Infinity]}]

categoryMembers[title_] :=
 Module[{url =
    "http://en.wikipedia.org/w/api.php?action=query&list=\
categorymembers&format=json&cmlimit=500&cmtitle=Category:" <>
     URLEncode@title},
  Join @@ NestWhileList[
     a@Import[
        url <> "&cmcontinue=" <> #[["query-continue",
           "categorymembers", "cmcontinue"]], "JSON"] &,
     a@Import[url, "JSON"], KeyExistsQ@"query-continue"][[All,
     "query", "categorymembers", All, "title"]]]
sample = {Hyperlink[#, "http://en.wikipedia.org/wiki/" <> #] & /@
   RandomSample[subjects, 3],
  Hyperlink[#,
     "http://tvtropes.org/pmwiki/pmwiki.php/Main/" <>
      StringReplace[#, Except@LetterCharacter -> ""]] & /@
   RandomSample[plots~Join~goals, 2]}
I got: {{Ibn Battuta,Cell membrane,Rowing (sport)},{Die Hard on an X ,Garage Sale }} Give me a week to churn on it.
 
hey what's up
anyone here?
belisarius???
 
@mildused yup
hi
 
Can you help me with something?
I have an image of sheet music and I'm trying to separate just the notes or just the staff lines. Any ideas?
 
@mildused No much more than a drunkard, but you've my sympathy
@mildused without using scissors?
 
3:36 AM
Okay that actually made me laugh out loud.
 
@mildused It doesn't sound too difficult, but ... why one could want to do that??
 
Okay let me tell you my story of why I'm doing the weirdest problems.
I'm 15 years old and used Mathematica to do simple planetary simulations for my Science Fair and won 1st place. Now I wanted to take my Mathematica knowledge to the next tier and compete in Intel Talent Search contest. To prep myself up, I bought a book full of exercises, but only has solutions to some difficult problems. I guess this isn't too difficult for the author. Whenever I have trouble, I seek help from my great friends on the InterWebs.
 
@mildused OK. So you're in a learning quest. Good.
Let's play Socrates!
 
yeah, and I'm having a BLAST so far. too bad school is starting monday
 
What have you already thought about that problem?
 
3:42 AM
I tried using MorphologicalComponents with Colorize
Can I send you the image so you know what I'm talking about?
 
@mildused Not necessary, I can imagine that. Morphological ops are needed indeed. But not enough. First go to Wikipedia and read about the "Hough Transform"
Then, go to Mathematica and read about "ImageLines[]"
Then, think about how to combine a Hough Transform with some Morphological ops
 
but the MorphologicalComponents with Colorize looks like the farthest thing from what I'm looking for
 
@mildused MorphologicalComponnents isn't a morph op. It just detects components!
 
oh sorry sorry
 
:)
@mildused Type for example ?Morpho*
 
3:49 AM
which one should I be looking at?
because sorry if I'm being dumb but I don't know anything above Calculus and some of those look pretty complicated
 
@mildused First check out the Hough Transform. That should light up some bulbs up there
 
I tried Morphological Perimeter but that looks awfully like EdgeDetect. Do those have similarities?

OK
 
@mildused One warning
 
Warning for what?
 
Reading that kind of things "lightly" i.e. trying to learn how to use it but not understanding how it works impedes you from achieving a superior mindset
 
3:55 AM
Ok Thanks!
 
"superior mindset" in the above sentence means "being a braggart"
 
well they deserve it don't they?
it's not easy
 
@mildused of course not. That's the only reason to study science.
 
you deserve to brag
 
@mildused 10^9 seconds of studying should earn you some privileges
 
4:02 AM
is that a typo?
 
@mildused no, just beer
 
thats like 30k hours?
 
@mildused one of those privileges is the one that allows you to exaggerate
 
Bro, I think you're drunk.
 
@mildused In any case I'm far worse when sober
 
4:06 AM
Well drink away I guess.
 
@mildused Go to your Hough transform business, and learn something for good!
 
I'm trying to understand but It's very confusing
 
@mildused Use Mathematica for learning.Draw a Line[] and program the Hough Transform by hand
 
Ummm Sure
I'll try that
 
@mildused BTW, you can check your program by using Mathmatica too Radon[ yourImage, Method -> "Hough"] // ImageAdjust will compute exactly that
 
4:19 AM
what exactly makes the function run for long?
like is it the scanning image?
 
@mildused Perhaps your image is too large ...
 
@belisarius not really
 
@mildused Also, be sure to work with binary (B&W) images.You don't need colors nor grayscales there
 
I can just Binarize[pic2] right?
 
@mildused sure
@mildused upload your image here
 
4:22 AM
Okay
 
binarized
 
Imgur?
 
@mildused just use the upload button next to the "send" button
 
dude, I'm either crazy or there isn't an upload button
 
@mildused PerhapsI installed some script. Use imgur
:)
 
4:26 AM
hmmm
cool
What I have: http://i.imgur.com/w6wWG6B.png?1
What I want: http://i.imgur.com/66nPpml.png?1
 
@mildused probably it's this one
97
Q: Can I easily post images to this site directly from Mathematica? (Yes!)

SzabolcsGraphics are tightly integrated into the Mathematica interface. The Front End is programmable, and Mathematica has functions to interface with the web, so the question naturally comes up: Could we make it possible to upload images to StackExchange directly from Mathematica, using a palette butt...

 
No it's not a script. It's a rep requirement.
I need 100
I think I got all 98 from 1 post though lol
But you're probably thinking "98! Hah. You think that's a lot."
 
@mildused Now you have'em
 
Hmm?
 
4:31 AM
Oh Dude Thanks so much
But I still don't see it
 
@mildused sometimes it takes a few mins.
 
no problem but did you see those images?
 
Try reloading the page every 5mins
 
I'll do that
It took a day to unlock the chat room after I had the requirements
 
@mildused lemme see
 
4:35 AM
And can I change that sound notification? It kills me
 
4:54 AM
@mildused This was my first try:
Closing[img, {Array[1 &, 10], Array[0 &, 10], Array[0 &, 10], Array[1 &, 10]}]
But you're trying to solve two problems here
one easy, one though
 
I think Closing and Opening and Erosion would remove details we don't want it to remove right?
What are the two problems?
 
the easy one is erasing the lines
 
Well guess what
 
the other one (much, much difficult) is dealing with a low quality image
 
Erasing the lines is pretty easy and then we can just use ImageDifference right?
so then we have the notes
 
4:57 AM
Try doing the same with a little very high quality image first
 
That is the only image givne.
given*.
 
@mildused That depends ... you need to preserve the intersections
 
Not really. I can have a crude remainder error.
But the better image the better learning teh better code
 
@mildused Anyway, that's a real life problem.Not one the author thought for you to learn. It's interesting, but doing it really well will take you a lot of time
with THAT kind of image
If you post the problem at the site,someone will solve it for you. But I don't think it as a standard procedure for learning
 
I guess.
 
5:03 AM
I think real life problems aren't a good path for learning. You need to pay too much attention to the details.
 
Well it is for Intel Talent Search preparation
 
@mildused I don't know it, sorry
 
It's okay
No need to be sorry
 
@mildused I'd just start with something really simple like this, and then add to it as you need depending on the range of images you need to process.
k = ConstantArray[1, {3, 9}] {1, .6, 1}; ImageFilter[
 If[Total[Abs[# - k], 2] < 3, 1, #[[2, 5]]] &, sheet, {1, 4}]
 
Wow! thanks
that looks pretty good
 
5:15 AM
I guess that just does the horizontal lines, but you could do another pass for the long vertical lines between the measures.
 
But that would remove the part of the ntoes
 
Not if you only removed vertical lines longer than the note necks.
 
hmmm
Well this the post I made. It should be more descriptive and have everything in one place. I will try what you said
0
Q: Removing Image Elements

mildusedI came across an exercise in one of the online tutorials for Mathematica and can't solve this Image Processing Problem. The files I reference to are attached. I have the Music Sheet (01Original.TIF) and am trying to produce the images 01Notes.TIF and 01StaffLine.TIF. I know that once I get eit...

 
 
4 hours later…
9:02 AM
@mildused you've probably figured it out by now, but you can change the sound notifications by clicking on the black speaker icon at the top of the banner to the right of this chat.
 
 
5 hours later…
2:19 PM
Wow. The tutorial video for the sheet music problem was much more thorough than I was expecting.
 
2:38 PM
I still haven't solved the sheet music problem
 
3:14 PM
anyone online?
@MichaelHale
 
Yes, I'm here
 
I was trying your algorithm for the image filter. and it seems to be taking forever
did it take a while for you as well?
 
Yes, ImageFilter can be slow, but it's pretty flexible for experimenting.
 
okay
 
3:39 PM
Could you try your algorithm with my image?
Please @MichaelHale
 
3:54 PM
@mildused While experimenting, try to crop out just a piece of the music sheet (unless you are already doing that.)
 
can you explain?
@Pickett
 
@mildused Well, you said it was slow. It's OK if it's slow and you only have to do it once but while you are experimenting you can take a part of the image that is not too large and quickly try different transforms on it.
I've been experimenting a bit on your images using the Radon transform and it's pretty fast because I only use a part of the image to experiment on.
 
@Pickett I'm only 15 years old. Can you dumb it down for me? You want me to take a piece of the picture. I can do ImageTake[]? and then just keep using that the whole time. I can ImageTake like the first line of the music?
Like this

ImageTake[original, {400, 1525}]
 
There is a built in tool for cropping images. Just select the image and then "crop tool" from the toolbar that appears underneath the image. This is the image I use to experiment on:
 
thanks!
I love how everyone is trying to help me.
This place is amazing.
 
4:09 PM
:)
 
@Pickett I'm out of ideas.
Possible to use Fourier Transform?
 
@mildused There are a few people on this site who are good at image processing. Hopefully they will take an interest in your question and answer it.
 
Hopefully.
 
@mildused My code didn't work very well on your image. I would copy the code in the video (it's a fair bit of code, but probably more robust and faster) or try removing peaks from the Fourier transform like someone else suggested.
 
4:25 PM
Did you try Removing Small Components on that Image?
Or some noise filter?
 
4:54 PM
No
 
 
3 hours later…
7:54 PM
@mildused check out the serious and not so serious image processing applications notebook found here wolfram.com/training/special-event/…
There is a section on optical music recognition
 
 
1 hour later…
9:11 PM
@Pato
@PatoCriollo I'm watching it right now. It's great but it didn't work on my example as well. I even decreased the threshold involving duplicate lines to overlay on the staff lines but some lines just don't pick up.
@PatoCriollo I was also a tad bit disappointed by Marilyn S. and Arnold M.
 

« first day (927 days earlier)      last day (3558 days later) »