June 1st, 2010
In case you wanted to test the automatic vectorization feature on clker, one possible way is to capture your own images using your digital camera and that’s a very easy process. However, a quick and easier solution would be to use a readily available public domain image. Images that are on the internet by default are copyright protected so you can’t grab an image from anywhere and “think” that it is public domain just because its owner did not say it is copyright protected. Some good sources of public domain images are:
1. My best source is wikimedia.org. They have around 47000 images that are public domain in the US in this category ( http://commons.wikimedia.org/wiki/Category:PD_US ). Most of those images comply with the upload policy of clker.com , some don’t.
2. The library of congress ( http://www.loc.gov/library/libarch-digital.html ). You have to be careful with that one and read the rights of every collection. Many images are dated before 1923, and most of those are public domain as long as they were actually published before 1923 ( you need to read the date on the description as well ).
3. US Government websites:
a. National cancer institue ( http://visualsonline.cancer.gov/ ): Some images here are in public domain, others are not so you need to read the rights under every image carefully.
b. US military websites: http://www.armymedicine.army.mil/news/photos/photos-new.cfm http://search.ahp.us.army.mil/search/images/?search= http://www.carlisle.army.mil/AHEC/mediagallery/photoGalleryList.cfm and lots of others. On google search for “site:.mil photo gallery” and you will get a lot of hits.
Generally for all gov and mil websites, if the picture is taken by a government photographer it is in public domain. However, if the image contains people currently alive, those people can request that their image be deleted. If the picture contains living or dead actors, then either them or their heirs can do the same. So if you are testing, keep your choices confined to photos of things and animals, and actually there are tons of those.
I also recommend renaming the files before uploading them with a more relevant title. Images with names SDC002123.jpg gives clker no clues about what the image contains, and does not help us to automatically classify the image. However, if the image is named giraffe_in_zoo.jpg that will actually help clker more to categorize the image.
Have fun vectorizing on clker.
May 28th, 2010
TAt clker.com, we have been adding newer tools that help artists as well as non-artist develop vector illustrations. The most recent tool that we added is a fully automated vectorization tool that vectorizes any high definition photo the user uploads. In our experiments, this automated vectorization tool produced the best results on images of objects with white background, or illustration images with few colors.
We will show a complete example of how to use the output of this tool together with inkscape (from inkscape.org) to produce an illustration image ( clip art ).
(c) Clip art
The original image that was vectorized is an image of a japanese woman sewing. The image was obtained from the library of congress (http://hdl.loc.gov/loc.pnp/jpd.02680). In Figure 1, picture (a) is the original picture. The original picture needs to be at least 0.7 mega pixel and the recommended size is 1 Mega pixel. If the picture resolution is higher than 1 Mega pixel, clker will reduce the size just during vectorization, and that helps with noise reduction.
Picture (b) is the automatically vectorized version. You will notice that when you click the SVG link the picture actually gets rendered in chrome/firefox/opera as SVG.
The third picture, which is the clip art illustration was obtained from (b) using inkscape. The picture was edited and the background was deleted. The Fluorescent color filter was applied to enhance the colors a little bit and the result was uploaded to clker. Producing (c) from (b) is a manual process because a human being is needed, who understands the picture.
(c) Clip art, cleaned with inkscape
Other examples include the “Lady talking in the phone” in Figure 2. The same exact process was applied. Picture (a) was uploaded with a resolution higher than 1 Mega pixel. After a little bit clker.com vectorized it and (b) was there. The SVG version of (b) was downloaded locally on my computer and edited with inkscape. All the objects around the lady were deleted and left the lady and the phone.
In both those examples the original picture was an illustration and the success rate when dealing with an illustration is higher than photographic images.
An example of a clip art that was generated from photographic image is shown in Figure 3.
(c) Clipart cleaned with inkscape
The same procedure was applied. However, one has to keep in mind that in photographic images there are lots of gradients, and so the edges of the vectorized version will not be crisp clear and the images might need more work to clean up compared to vectorized illustrations.
May 7th, 2010
Clker will be down approx. at 8pm eastern time for around 2 – 3 hours for maintenance.
April 30th, 2010
Automatic vectorization is up and working. Upload an image of at least 0.5 megapixel and after a little while you should see a vectorized version in your profile.
I got best results with images larger than 1 megapixel. The website service will automatically resize it in case it was larger than 1MP down to 1MP. Try to edit the picture before uploading and add any effects you want like an oval or heart shaped crop.
Soon, I will add another interface by which users can request custom trace or to retrace an image that has been already vectorized before.
It already started generating some pictures, and I already see some errors esp. with alpha channel png files, so stick with full color ones for now . I’ll go to sleep now, and tomorrow I will write a detailed post.
Send feedback whenever possible.
April 25th, 2010
I tried ubuntu lucid lynx (10.04) on my laptop and my first impressions it’s way better than 9.10. I have 9.10 now running on my desktop, and tried it on my laptop as well. When I tried to install 9.10 on my new laptop (hp dv4 series) the screen didn’t show anything. On my desktop, compiz does not work on 9.10. When I tried 10.04 it worked and all the hardware on my laptop worked as well. Still to try it on the desktop machine.
Other than the screen, the new gnome theme is very slick and together with compiz they are really way slicker than win7, which runs on the same laptop. I installed all my work on the ubuntu installation including gcc, emacs, vim …etc.
The installation was very smooth. I tried to look around on wubi’s website for a way to install 10.04 but I didn’t find. After some googling I found a link on Librarian Geek’s website.
Some draw backs: 1. happycoders emacs is no longer maintained. Ubuntu removed it from the release. 2. Flash 64 bit does not install correctly from adobe’s website. I got it to work by downloading the tar.gz file pointed to by Raman here. The one from adobe for some reason does not work, and the channel does not work as well.
Given it’s alpha status, I think it will be an excellent LTS release for ubuntu. For me, it’s a keeper on my laptop.
April 23rd, 2010
SI had some delays due to unscheduled projects that were done in my home. I’m just starting to gain control of my time again, together with the schedule of contractors. That pushed my two weeks off. Now, the tracer can produce outlines. They are not amazing, but some times they enhance the traced image. An example is this hot dog sandwich:
This is one example that I think edges produced good results. The thickness of the edge will vary with its position and my objective was to make it appear as brush strokes, which are thin at the start, and tapers off at the end of the line. Other examples which I think that it did not produce good results is the Hillary Clinton image.
So it seems that with simple images, the outline will produce better results, while with complicated images it will not. I will start integrating the tracer with clker form today, and will default its operation to trace without producing outlines. We will also allow every user to change their default behavior for the tracer. Watch for another blog post withing the next couple of days.
April 7th, 2010
Automatic tracing of uploaded photos is one important feature coming to clker.com . I will describe briefly how far we reached, and what I hope to accomplish within the next few days until it is released.
First off, let me outline the benefits and draw backs of auto tracing. The obvious biggest benefit is no human work is put into the segmentation and vectorization process. However, it has been proven in lots of domains that when two humans segment the same picture their result will have approx. 80-85% common areas. This means that humans will disagree about the segmentation of the same image in about 15% of the image itself. That’s because the way our brain interprets an image depends on every persons own understanding and knowledge of the contents image, and that can vary a lot. Even when the human segmentation results are in a very specific context like brain segmentation, and done by radiologists the comparison between the similarity between their results was between 83-87% using a similarity measure similar to the Jaccard index. So the disadvantage of automatic segmentation is that if we were able to reach the perfect algorithm to segment an image, the user will always have around 15-20% disagreement with its result.
The results that I will show here are based on this helicopter image (also shown on the top). There are several parameters used as the inputs to the segmentation program. Most of those will either be selected by us to provide good results in average, or passed through clker’s interface in a more humanly understandable fashion. The image was reduced in size to approximately 1 mega pixel before processing.
Those two images are direct output of the vectorizing program without any modifications. The difference between them is the number of regions in each. In the image on the right, we attempted to reduce the number of resulting segments, which can help greatly especially if the user plans on editing the image after.
To see how hard it is to produce a clip art out of this segmentation, I tried to edit the image on the left (the one with larger number of segments) and delete all the sky regions. It took me around a minute to do so, just select and delete and the result is below:
However, some people will not perceive this image as a clip art. One reason is the lack of hard outlines around the object when compared to an image like the deck officer here . The other reason is the number of regions is way smaller. I will be spending the next couple of days trying to get some borders up and hopefully we’ll get some good results.
I imagine now two ways to have this feature up. The first one once you upload a raster image, clker will work on vectorizing it then you will get an email with a link to the vectorized version. The other way (which will come later in time) will be requesting a custom segmentation with the ability to adjust some of the parameters like for example requesting smaller number of regions, or different types of color enhancement including equalization, white balance correction ..etc.
Hopefully this will be running on clker by Monday next week.
February 19th, 2010
We released a new clker interface on Wed. The new interface has white as it’s main background less colors with blue being the main focus. Some of the elements & buttons are currently being worked on to follow the same theme. The new interface is not colorish like the old one, as we learned from users that too many colors become distracting and interfere with the images that are already colored.
We redesigned the home page and reduced tons of clutter. An improved version will be released soon that will give larger space for the community updates.
I would like to thank both sciweavers.org & Florian with Das Elektronik Lexikon for valuable insights and reviews during the process.
Keep your eyes on clker for new features released within the next couple of weeks. Those are definitely worth watching (I’m not going to say what they are now ).
February 3rd, 2010
When I was studying for my PhD at the University of Miami, I had Fedora installed on my machine as the primary os and switched to windows 2000 whenever I was forced to. I remember that was about 2002 / 2003 and I was able to get all my work done over there including posting lab notes to students, writing my papers and research reports on tetex and of course writing code. I don’t recall that octave was there, so I had to use windows to run simple experiments on Matlab.
Since I didn’t have a desktop machine, I decided to setup one. We had an old dell optiplex GX260 with an XP license and of course no media, which was bought within a lot of machines and so no hope to actually obtain the original media i.e. the license is just useless. The OS on the machine was totally damaged beyond repair.
I decided to install Ubuntu desktop and I tried both 8.04 and 9.10 (sorry I have hard time keeping track of the animal name). Amazingly, 8.04 flawlessly worked but 9.10 compiz had an issue with the on-board graphics card. I installed google chrome, flash player (which are not part of the ubuntu repository), and currently using gmail chat instead of gtalk. Hopefully I will try to setup pidgin to be able to chat in voice.
I always knew that I don’t need MS windows. My MS Windows is used to run openoffice.org and putty 90% of the time!!! Hopefully I will find a good graphics card compatible with 9.10 on newegg and will upgrade soon.
January 28th, 2010
I’ve made the decision this week to work full time on clker.com . So clker.com, which was my hobby website will now get at least 10 times more attention and time from me.
How does that affect clker’s users? You will see faster improvements and updates – and may be more radical changes to Crayon and other new components. Make sure you file all bugs and feature requests here, which will greatly help me plan the releases.