Ontology2 announces the beta release of the Ookaboo RDF Dump, which contains metadata for nearly 1,000,000 public domain and Creative Commons images of more than 500,000 specific topics from Dbpedia and Freebase.
The Ookaboo RDF dump is released under a CC-BY-SA license that is friendly to both academic and commericial use. With precision in excess of 0.98, Ookaboo enables entirely new applications for image search and classification.
We started Ookaboo because we weren't happy with the state of the art in free photograph collections. Although we've improved on the accuracy of image search, so far, we've done a poor job of organizing topics in our browsing interface. This is a consequence of choices we've made: to scale our process to 500,000 topics, we made no effort to organize the topics before inserting them.
We've rolled out the first software upgrade in a series that will improve navigation in Ookaboo. Initially we're tackling the problem of topics that are categories, such as "City", "Car", "Person" and "Animal". An example of a new page can be found here: Pictures of cities.
We hear from people every day who'd like to understand how to use free pictures from Ookaboo. Pictures from Ookaboo can be used freely, but you must you them consistently with the law -- this page explains how to do so.
Ontology2 founder Paul Houle spoke at the NYC Semantic Web Meetup last week. His talk, "What Would You Do With Free Pictures of Everything on Earth?" described the origins of Ookaboo, the technology behind it, and the API that makes it possible to find pictures using precise Linked Data Terms. The slides to the talk are available on slideshare.
The talk tells the story of three predecessor sites to Ookaboo, animalphotos.info, carpictures.cc and ny-pictures.com. By expanding Ookaboo's scope, we were able to, paradoxically, simplify the process of taxonomy construction. This kept us focused on the process of accurately attaching pictures to topics. The talk invited the audience to "build their own Linked Data cloud" using hundreds of data sources that can be linked to Freebase and DBpedia and, by doing so, creating entirely new kinds of applications, such as the Ookaboo API.
Creative works produced by employees of the U.S. government are in the public domain. As a result, many thousands of pictures shot by military photographers are in the public domain. They document everything from the Los Angeles River to Afghan bread, and they give us a glimpse of the world of our fighting men and women.
Anyone who takes great pictures and makes them available is a hero for us, but military photographs from Ookaboo reminds us of our soldiers in the field, their families, and those men and women who have served in the past. We thank them all and we thank the soldier's media center for this image.
The same public domain policy means the world has access to photos from NASA, the Department of Agriculture and many other government agencies. Geographic and census data are available for free here, a situation quite different from Canada, on our northern border. The U.S. has done many things right and wrong, but free access to works created by our tax dollars is something to be proud of.
About two years ago I rediscovered the Japanese Pop Group Shonen Knife, and was listening to the chorus of the song Riding the Rocket. There was a word that got repeated, and it took me several attempts to find it in a Japanese-to-English dictionary. 
The word turned out to be 浮かぶ which is commonly transliterated as "ukabu" which means "floating"; once I discovered this word, I heard it being used a lot in spoken Japanese, and saw it could be used in the following contexts:
An astronaut weightless in space
Floating in a pool
The feeling one might have dancing or jumping on a trampoline
The feeling one might have while daydreaming or on the edge of sleep
As of November, Ookaboo contains 941,112 pictures of 517,089 different topics that were collected with the help of 171,654 different sources.
Last month we implemented a new metadata ingestion module in response to a problem spotted by a user. We used this to correct the attribution records for about 1500 images. Although our metadata ingestion process is highly accurate, we depend on the vigilance of Ookaboo's users to help us find and correct problems. The new metadata ingestion module is simpler and more flexible than the old module, and may soon be used to improve other data records and for the ingestion of new images.
Ookaboo's semantic technology, remarkably, has allowed us to amass nearly a million well-organized images with a minimum of human involvement. Ookaboo has a critical mass of content, and is beginning to attract a significant audience. We're on the threshold of developing community features that will realize Web 3.0.
Ookaboo is the latest in a series of sites developed by Ontology2 and we've always had a product line strategy that would require federated login. For a long time we've considered OpenId, Facebook Connect, and developing our own in-house system. In that time, Facebook's offerings have become increasing attractive.