|
Wheany posted:Content aware scaling/liquid scaling Not quite as cool; I think I did a little too much content aware scaling. I also loaded the pictures in reverse which is why it makes things seems to linger. Definitely going to poke around with this.
|
# ¿ Apr 20, 2014 00:06 |
|
|
# ¿ Apr 28, 2024 05:48 |
|
Figured out that using protection during content aware scaling can preserve the loop in gifs. Could be used to preserve other parts of time too I supposed.
|
# ¿ Apr 20, 2014 06:33 |
|
If anyone else wants to join in, this is the script I wrote for processing: Files are expected to be in png format with 3 digit numbering, stored in a folder called data. code:
Wheany posted:Have you ever content-aware scaled... different color channels... temporally? How did I not think of this? So far I've tried motion blur, edge detection, and a few other filters, but nothing has been too interesting. TheLastManStanding fucked around with this message at 18:59 on Apr 24, 2014 |
# ¿ Apr 24, 2014 18:57 |
|
Some more scripts for people to use. Splits frames into red, green, and blue frames. code:
code:
Smearing: this expects a background image numbered '000' to compare against code:
|
# ¿ Apr 26, 2014 03:13 |
|
PHIZ KALIFA posted:So, I've got a question for folks about a potential Processing project: I'm looking for something I can feed a video clip to that will break it into 10-15 second chunks, rearrange it, and pass a filter over what results. Would that be feasible? My background in coding is I know how to make TI's draw ding-a-lings on the graph page. I like this one.
|
# ¿ Oct 6, 2014 05:12 |
|
Wrote a pair of programs yesterday to help in glitching pics. The first does pixel sorting; it can load pixels by RGB or HSB, can sort individual channels, and can swap and recombine in both RBG or HSB. Sorting seems to work best on pictures with large areas of similar colors; pictures with lots of colors tend to go muddy. Cross processing from RGB to HSB tends to give good results. You can also mix input channels.code:
|
# ¿ Oct 7, 2014 23:51 |
|
The second is interactive. The mouse wheel is used to scroll through tools (What you're using is shown in the console), left click to use them, press 's' to save the picture. There's various line tools, cut and paste, inverting sections, posterizing sections, circles, and an eraser.code:
TheLastManStanding fucked around with this message at 05:09 on Oct 10, 2014 |
# ¿ Oct 7, 2014 23:56 |
|
Updated my code; added the ability to push and pull the image, as well as stripping color channels and adding noise. - Arrow keys control line direction - r, g, b, & a keys select which channel is to be modified (a is all) - [ and ] resize the tools - The eraser is now right click - mouse wheel changes tools - s to save code:
|
# ¿ Oct 10, 2014 05:08 |
|
Solumin posted:You should really throw both of these tools in a Github repository! I seriously hate the github interface, but I ended up making one: https://github.com/Phonocardiogram/imageTools/ Tell me if that works. I included a new version of my PixelSorter, which has a lot more options and shows an immediate result. I also updated GlitchPaint. The changes are: - Right click now uses a brush to erase - Greatly improved the noise math - Changed how cut/paste worked - Added a quick save/load feature so you can test out small changes easily - Added 5 filters (Horizontal/Vertical blinds, quilting, random/uniform pointilisim) which are also affected by brush size + If anyone has suggestions for filters I'll see about adding them. Right now the only thing I can think of is a harsh jpg compression style filter.
|
# ¿ Oct 12, 2014 08:11 |
|
https://github.com/Phonocardiogram/imageTools/tree/master/TimeScaler Added a new tool. This one performs a content-aware-scale directly on the time dimension of a sequence of images. This cuts out the intermediate steps of slitscanning and using photoshop's content aware scale, and it eliminates the horizontal artifacts that the previous method produced. It also allows for more direct tuning of the intensity of the effect and greatly reduces the time required to get a result. The only side effect is that the math on this one melted my brain These two gifs show how the results can be tweaked.
|
# ¿ Oct 15, 2014 10:50 |
|
Max posted:There are people in this thread that absolutely must submit to this. The Tate is asking for people to openly glitch their art to display. I changed my time scale code to properly interpolate between frames, but the difference is almost unnoticeable. It got rid of hard pixel boundaries on quick changes, but also introduced a weird temporal artifact in which contrasted outlines from later frames are apparent in much earlier frames. I haven't updated the code on github yet since I plan to completely rework the order of operations as I'm not satisfied with how smooth it is yet. But have some videos: https://www.youtube.com/watch?v=tRadBW9ZWNU https://www.youtube.com/watch?v=VjZe_hsnMa0 Row sorting RGB: Row sorting hue and sat: Full sort red, row sort blue: Progressively smaller quilting, then vertical blinds:
|
# ¿ Oct 24, 2014 00:58 |
|
Updates for https://github.com/Phonocardiogram/imageTools GlitchPaint - Now has a recording feature, toggle with the ` key. It saves every 5th frame just so you don't eat up space too fast. You can change that number if you want. - Group pixel sorting, both as a tool that can be applied and as a filter. - Scalable cut and paste as well as being able to cut/paste individual channels in both RGB and HSB. TimeScaler - Reordered things. It should give a better/smoother result on fast moving scenes. Haven't thoroughly tested it yet though. PixelSorter - Added group pixel sorting and tweaked a few things FrameTools - New tool which repeatedly applies a filter to a single frame and outputs each successive addition. Mostly fluff, but mode 7 (dissolves images using small group sorting) and mode 5 (repeated quilting) are really fun to watch.
|
# ¿ Nov 3, 2014 06:13 |
|
Now that's the stuff.
|
# ¿ Nov 10, 2014 21:29 |
|
You need Processing
|
# ¿ Dec 6, 2014 06:14 |
|
That's HTML. It shouldn't look anything like that. Did you download as a zip? I'm guessing you copy pasted it wrong.
|
# ¿ Dec 6, 2014 07:05 |
|
Steak posted:I can't for the life of me get any of Last Man Standing's scripts to work or do anything. Possibly related: my coding knowlege extends as far as copy and pasting code into notepad to make scripts for Arma.
|
# ¿ Dec 19, 2014 10:28 |
|
Update to GlitchPaint Now with a fancy GUI Other changes: I integrated PixelSorter, so you don't have to switch back and forth. The line tool can now be used at any angle. The noise level of the noise tool is now adjustable. Paste tool has a preview. I made a very basic glitch filter. I also added a filter which reduces the color palette with a floyd dither. The options are a 1-bit black/white, 3-bit 4-bit 8-bit RGB, 4-bit Gameboy, NES, and an optional custom palette space. When my glitch filter glitched out: What value overflow does to a dither algorithm. Random other
|
# ¿ Jan 1, 2015 06:44 |
|
This is awesome. Is it a rotating brush or some mathematical squiggle used as a mask? I also like this and it should be super simple to put into GlitchPaint. Do you have any feedback on the new interface? Is it straightforward enough? Anything you'd like to see changed or added?
|
# ¿ Jan 3, 2015 22:16 |
|
Read the shortcuts tab or text file. '[' and ']' change the brush size by 5. '-' and '+' change it by 1. By holding the key you can go from size 1 to 100 in about a second. The mouse wheel will scroll through the tools and tab will change the mode.
|
# ¿ Jan 4, 2015 01:22 |
|
Another Update: (Download) I've given up on github so it's a zip file with the option of running the program as a windows 64 application or running the source in processing. Changes made: - The window is resizeable allowing you to maximize the window. - Added halftones in both black/white and color (it's not a true color halftone, but I might do that later). - The amount of dither is now controllable. Negative mean less dither, positive causes it to overcompensate which leads to interesting effects. - Mosaic with an optional gap. - Triangles, both uniform and random. - Blinds can be inserted manually and can be offset. - The sort brush now works as a selection.
|
# ¿ Jan 16, 2015 05:00 |
|
New Program: (Download) Had some free time yesterday so I threw together a RuttEtra style program. - Full 360 rotation along the X and Y axis. - X and Y resolution control. (Warning: Using very low x/y spacing on large images can pretty much stall your computer. When working with large images, slowly lower the spacing. I might limit this later.) - Line thicknesses. - Displacement mapping using any combination of RGB and/or HSB. - Perspective or Orthographic projections. - Display results as a single plane, horizontal lines, floating pixels, or a mesh. - It also lets you load a custom displacement map for added fun. - As always, the 's' key saves the frame. Example of the different modes: Example of line thickness at high line spacing: I've found that looking behind the images often gives the best results:
|
# ¿ Jan 19, 2015 07:33 |
|
Nope. Is your java up to date? I'm assuming that was an attempt at running the applet and not the source code? The applet only works for 64-bit Windows, so if your on a mac or something else you'll have to install processing and run the source (make sure to rename the source folder to match the program name).
|
# ¿ Jan 31, 2015 23:38 |
|
AphexMandelbrot posted:Take picture. quote:President Kucinich posted:Makin diamond flowers.
|
# ¿ Feb 5, 2015 22:54 |
|
It's a Delaunay Filter, most likely this one. If you do use it you'll want to add the lines code:
code:
code:
|
# ¿ Feb 12, 2015 01:19 |
|
President Kucinich posted:I like this one. It runs like poo poo, and constantly stalls out, took forever to get running right, but it's got some neat options in it. code:
TheLastManStanding fucked around with this message at 05:41 on Feb 12, 2015 |
# ¿ Feb 12, 2015 05:34 |
|
I might get around to finishing it. Most of the code is there, the thing was I figured if I was going to finish it then I wanted it to have at least one major thing that DMesh doesn't have, of which the only things I that came to mind were point-dragging and manual color application. The latter is easy, but the former is tricky. Moving the point is super easy (It's probably two lines of code), but doing it efficiently is tough since in most code it requires triangulating the whole mesh every frame (while dragged), which is exceptionally slow with only a few hundred triangles. My thought was to only re-triangulate the faces connected to the moving point (and the faces connected to those), but proper re-triangulation would require the total area of them to be convex, which becomes more unlikely as you add points. My current thought is to do something similar, remap a small section around the dragged portion, then delete the exterior faces. I'm thinking that it will be good enough in almost every case, and even when it does mess up it won't be an issue as the whole mesh will be recalculated when the mouse is released. I'll have to think about it some more though. In the mean time.... New Program: (Download) Takes an image and lets you slice up the polygons. It has various blend modes, deletable faces, random slopes, undo up to your last 100 actions, full stroke control, and auto points using two point edge detection (because of geometry, the results will be more abstract than you might anticipate). As usual, the app is for win-64 with Java 7, source included for everyone else.
|
# ¿ Feb 13, 2015 09:08 |
|
Also some junk I did in Re (which I've been meaning to update).
|
# ¿ Feb 13, 2015 09:11 |
|
Update: Re Version 2 (Download) - Updated the menu to the new style - Added blocks and vertical line modes - Changed how RGB displacement is calculated (looks much better now). - Offloaded the displacement to some custom shaders That last bit means that the shape only has to be remade when switching modes or resolution, so changing line thickness or displacement are instantaneous. The side effect of this is I had to change how the mask worked: Displacement is now based off of both the RGB/Brightness of the main image and the mask brightness, with the upside that they are all individually scalable. It also caused a minor change in how lines are rendered, but it shouldn't be an issue.
|
# ¿ Feb 17, 2015 03:30 |
|
Odd. If you're running the app you shouldn't have to do anything. I downloaded it and it's working for me, so the link is fine. Could be an OpenGL/video drivers issue. Were you able to get the other program running (slicer)? If you copy the data folder into the source folder and rename the source folder to Re you could try running it in Processing, which should give an error code if there's a problem. As a side note, Processing runs on Java 7. The current Java is version 8 and if you update to it it will suggest that you uninstall old versions, which means Processing apps won't run, so you should make sure the version you have is the newest version of 7. (You can have both versions installed as they are independent. Java is silly like that). TheLastManStanding fucked around with this message at 08:43 on Feb 18, 2015 |
# ¿ Feb 18, 2015 08:38 |
|
Try replacing the data folder with this one
|
# ¿ Feb 19, 2015 08:08 |
|
Sooo... Two weeks ago I finally got around to starting on my Deluanay program. I started by poking around other peoples code to see how they were doing it. Nearly everyone was using some slightly modified form of the triangulate library, which seemed convenient: However, in my tests it sometimes gave improper triangulations (for 5000 points there was around a 1/10 change of a mistake, typically a small and almost unnoticeable, but still a mistake). After a week of poking around the code I wasn't able to completely fix it, got frustrated, and threw everything out. I took a few days off from the project and when I came back I decided to look for a different route. I found one person who wasn't using the standard code. I tried it out and while it had some errors and was significantly slower (4000ms vs 600ms for 5000 points), the code was much simpler. I started with fixing the circumcircle equation and switched the inTriangle function from the dot product method to ray casting; neither gave improved speed, but it fixed the errors. Sorting the points gave a good boost bringing the time down to 1000ms. Still slow, but I figured it was decent. I then tweaked everything line by line, completely rewrote the face flipping and adjacent finding algorithms, and switched to a TreeSet to remove coincident points. That brought the time down to 50ms Somehow I had written a Deluanay triangulation program was 10 times faster than the one everyone else is using. Anyways, have fun with it. I'll probably clean up the triangulation code a bit more and release it as a library. I also have some more features I might try out. New Program: (Download) It makes Deluanay triangulations. Add points, delete points, move points, points that make holes, random points, points that use edge detection, plus all the overlay, blending, and stroke options from my other programs, plus a few more. You can use the a/d/c/h keys to quickly change modes. Right click a point to delete it. Play around with everything else.
|
# ¿ Mar 1, 2015 10:03 |
|
|
# ¿ Apr 28, 2024 05:48 |
|
You should be using Java 7u75 and make sure your graphics driver is up to date. If you're running it through processing and it throws an error I might be able to tell you what it means.
|
# ¿ Mar 4, 2015 08:37 |