Slashdot Mirror
PhotoSketch Image Manipulation Tool Taking the World by Storm
PhotoSketch, a new image manipulation program that combines stick-figure sketches, internet image search and pattern matching, seems to be spreading like wildfire. Created by five Chinese students at Tsinghua University and the National University of Singapore, the tool takes a basic sketch and simple labels and turns it into a polished image. "Although online image search generates many inappropriate results, our system is able to automatically select suitable photographs to generate a high quality composition, using a filtering scheme to exclude undesirable images," say the PhotoSketch team in an abstract outlining the tool. "We also provide a novel image blending algorithm to allow seamless image composition. Each blending result is given a numeric score, allowing us to find an optimal combination of discovered images. Experimental results show the method is very successful."
Since the link to homepage in the article is some old-dated one, here's a correct one:
And the binaries (it's a few command line programs, so no fancy UI)
This image looks sketched. I can tell from a few of the pixels, and from having seen a few sketches in my time.
The authors of the program--Tao Chen, Ming-Ming Cheng, Ping Tan, Ariel Shamir, and Shi-Min Hu at the Department of Computer Science and Technology, Tsinghua University, and the National University of Singapore--presented it at Siggraph Asia 2009.
An event that will be remembered forever in the History of Humanity as the day in which a million dorks were finally able to put themselves in X-rated positions with Megan Fox.
Did you ever wake up in the morning, with a Zombie Woof behind your eyes? -- FZ
Now NO ONE will believe the pics of me with Jessica Biel, Kate Beckinsale and Dolly Madison are real!
Populus vult decipi, ergo decipiatur...
"Force shits upon Reason's back." - Poor Richard's Almanac
This will make things way easier for Iran and North Korea.
I tried to draw a picture of a man with an erection. I labeled him "porn guy". Then I drew a picture of a woman with her mouth open and labeled her "porn whore cumshot".
The composite picture was fine except that the man and the woman were far apart from each other. In addition, even if I were to draw them closer together (hey, I'm working with a mouse here), the result would still have been sized incorrectly.
This technology holds lots of promise and is already pretty cool. I hope they can work out the kinks.
The reality is that it was only a matter of time before someone came up with something like this, with examples like Microsoft Photosynth, but this is an unbelievable implementation.
I'm not 100% sure, but I can definitely see the potential for Google to snatch this up really fast and incorporate it into Picasa or even google image search or something. The fact that something like this allows anyone (not just artists) to come up with novel images with minimal effort is fantastic. I do wonder how canned the images were though. IE: did they GIS for an image first, then use the image as a basis to draw the stick figure, knowing that their algorithm would pick the image they selected in the first place? I would like to see a live demo with an unplanned audience member doing the drawing. Then I'll really be impressed.
In related news anyone supporting current copyright laws have reinvigorated the economy after having to go out and purchase new pants. Cue the next great debate about copyright as we continue to try to shoe horn old ideas into the new world.
If you tagged the big circle with distended anus, then probably so.
Soon I can write a story and then I just compile it and it will show sniplets of existing movies or rendered characters and woha it's converted to a real movie even with end credits: Directed and written by ME ME
Oh I can't wait.
It appears from the video that it's running a fairly sophisticated series of algorithms to compare backgrounds and determine how difficult it would be to do a convincing mask-out of the foreground object, of which it appears to have a sort of heuristic expectation of shape from the user's sketch.
For instance, if your background is a grassy field, the user has requested a dog running, and you have a photo of a dog running over grass and a dog running over pavement, the grass one will allow a greater margin of error in the masking and thus it gets selected.
Overall, this looks like a fantastic step forward for computer vision, bringing the computer ever closer to the non-Cartesian way our brains see.
Your mind is clear / The things that you fear / Will fade with how much you / Believe what you hear
Having just taken a quick look through the config files and readme from the binary.zip file, it's pretty obvious this is very much a Proof of Concept release. You need to hard-code the number of sketched items, label them each in the config file, download the potential matched images to a specified directory, etc. It involves enough guess-work and too little documentation for me to proceed further, which is unfortunate. Has anyone else actually gotten it to work as described to confirm it does what it claims it can?
Anybody else see what's wrong with this picture?
"I'd rather be a lightning rod than a seismometer." -Ken Kesey
An interesting point: This research is being done in China, not the United States. Whatever happened to basic research being done in the US? Today's PARC laboratory is not in the US, but appears is in China.
This is not a good thing for people who live in the US. America's increasing dependence on outsourcing is destroying the US' capability to be competitive in today's environment.
The Harvard Business Review has an excellent article about how America is destorying its own future.
MaraDNS is an open-source DNS server.
Wow. I'm instantly instilled with the urge to plug an XKCD comic into this and see what happens.