WHAT IS THIS?
Videorooter created a map of available openly licensed videos. As part of this, we:
- Provided a draft open API for developers to use. Look up videos through the API and make use of the basic information provided — the title, description, author, license, and url — in your own applications.
- Tested and developed an algorithm for video fingerprinting. Our best algorithm had a 79,48% accuracy (under certain conditions!) and we know it can be improved! You can fork our Github repository, and it will automatically include any changes to your fork in our testing.
If you’re interested in our algorithm work, here’s what you need to look at:
- Our blog post about the testing environment
- Our statistics page which give an overview of all Github repositories tested
- Our algorithm repository template has some information about the configuration file needed in your Github repository
- The testing environment repo has a Dockerfile which show what kind of environment you can expect your code to be compiled in on our side
GETTING STARTED WITH THE API
# This will get you a random video from our current database, returning the information # according to the API specification. Current data is taken from Europeana and includes # about 3500 videos. $ curl -L http://devc.commonsmachinery.se:8080/random?type=video # If you have used our experimental blockhash to calculate a hash for a video, for instance # the video "Oosterschelde wordt afgesloten", you can see if Videorooter can find the video only # based on its blockhash: $ curl http://devc.commonsmachinery.se:8080/lookup/video?hash=3f017f0037c0e9c03c38383c7870fe00007f1cd484f1e8383e187f007e087e08
The returned information should include a distance metric (the smaller the number, the closer the match), and a reference you can look up in turn to get the metadata about the work from the database. If you want to explore further, you should have a look at the API specification, as well as follow along in our Getting started with Videorooter API blogpost.