If PhotoSketch works it will change the way you work
This exciting application has been called mindblowing and better then sliced bread. It was developed by five Chinese computer science and technology students at Tsinghua University and the national University of Singapore. It could be highly significant for Photoshop users and Stock Photography users and suppliers. If applied to Stock Photography a number of copyright issues would have to be taken into consideration though. For now, we’ll keep a close eye on the news about this project.
PhotoSketch, (the internet image montage ) project creates images from a combination of image recognition and keywording and creates an image from a rough sketch with these elements. It works like this:
The application has allready received a lot of attention and reactions on Mashable, Gizmodo, and ZDnet. Due to the potentially revolutionary consequences of this application not everyone believes in it being genuine, this was not helped by the site going down as a result of the Global interest.
Let us know what you think and what the consequences of this application would be…if it works
It would be amazing if it works. Would be a game-changer if it worked WELL, if you follow.
But yes, the copyright issues are formidable….
this software would certainly bring copyright infringement to the forefront. it also could make it more difficult for a photographer to even recognize a portion/s of their image that may have been used to create the final composite image. it unfortunate;y could be a mega boom for stock photo houses worldwide with shared copyright owned images. it would change the way photographers work, we’d be in the soup line.
buy shares in stock photo business’
Not impressed. Not photography.
If PhotoSketch works it will change the way you work forever! | http://www.fastmediamagazine.com/?p=1052 (internet image montages) =]
Hi,
Nice results, but if you look at their paper you can see that it takes 20 minutes per image even when using multiple fast PCs, so I don’t think this is close to being viable for real use yet.
Rgds, Joe
As a result of the great interest in the PhotoSketch article I asked Tony Rowland of Imense to give me his perspective. This is what he sent:
It looks like an interesting method of using computer vision for image/design composition, rather than image classification and retrieval which is where imense are pioneers”
Sketch-based query interfaces have been around for over a decade, for example the Berkeley “Blobworld” system of the late 90s:
http://aa-lab.cs.uu.nl/cbirsurvey/cbir-survey/node9.html
http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.46.6474
Cambridge Ontology had sketch based CBIR back in 2001. The problem is that people are not willing to invest a lot of effort into sketching a query, and object outlines vary greatly based on viewing angle, occlusions, lighting etc., so this kind of shape matching is very brittle.
For those reasons, PhotoSketch are only doing text-based image search, since the user has to label each sketched object outline so that the system can find candidate images based on tags. So there’s no object recognition or classification or Ontologies whatsoever involved in this.
It looks like a great project though and we would love to see them test on more than a few thousand images and successfully reduce processing time per image down from 20 minutes.