While everyone is searching for what commercial stock might become in the future, the Chinese might have one solution and it is quite revolutionary. Researchers from TNList, Department of Computer Science and Technology, Tsinghua University along with a Singapore based researcher, have put together Photosketcher, previously called Sketch2Photo.
The principle is quite simple : you sketch what and how your final image image should look like and it fetches all the elements on the internet and stitches them together for a final composite image. Better yet, here is a visual explanation :
This would be a major breakthrough for art directors all around the world. Instead of going to hundreds of stock banks trying to locate an image that remotely looks like the one they had in mind, they could actually build it from different bits and pieces of a variety of images. Obviously, they wouldn’t even have to hire photographers for photo shoots anymore, at least for those that have common elements in them.
But, before we get there, this is a pretty nifty tool for art directors who would like to create comps to show their clients or the photogrpahers they are about to hire what they have in mind.
In the paper, it is unclear where in the internet the search is performed. It seems to be everywhere, creating a huge copyright headache. Of course, some would argue that since the end result is a composite and thus a new creation, there is no copyright due. Let’s leave the arguing to Lawheads and revue the implication of this new tool.
If the search was only applied to Creative Common content ( Flickr) one could probably be free of any copyright issue. Furthermore, one would have the legitimate right to register the new finished product as their own and license it. Think about it. A completely computer generated image created with bits and pieces of images from various photographers would come and take its rightful place next to work from long-time pros. Wow. And some though microstock was bad. Wait until everyone can create photographs.
Currently, the end result is very average as the selection of images does not pay any attention to light orientation and shadows. However, that could easily be an additional search parameter which would allow for extremely realistic end photographs.
What would this imply for editorial photography, especially news ? Major, major trouble. On could easily put together, in any environment, two people that have never met and look very realistic. Our news imagery could suddenly be flooded with hand-made images of events that have never taken place. Would we ever trust photography ever again ? Doubtful. Photography will have to go through rigorous credibility checking before being branded as real news.
Finally, could this Photosketcher be a hoax ? Doubtful. Finding image via sketching is already widely operational, while automated extracting already exist ( Adobe has a great one in Photoshop Element). Stitching, as we all know, is also very common. Thus combining all these know application together is not impossible. It is actually not too hard. The whole operation must take a pretty hefty amount of processing power but then we have no information on what type of computer these students have.
This new tool, however amazing it seems to be, has many implication for the world of photogrpahy and will have far reaching repercussions. It’s acceptance and usage will be something to monitor closely for anyone involved in photogrpahy.
See a full explanation in this video:
Sketch2Photo: Internet Image Montage from Tao Chen on Vimeo.