Search This Blog

Friday, 3 June 2016

Vitrivr is an open source engine that lets you search for videos with a sketch

The technology is currently in development, and freely available on GitHub. In a time where we carry GB’s worth of photos and videos on our smartphones, it’s no wonder a number of us will have ponderously massive collections of videos stored at home. Which is all well and good, you have everything you want to save tucked away somewhere, but how do you find any of it?
Even major companies Google and Apple have recognised the issue at hand, implementing new sorting and search functions in their local and cloud storage tools. But researchers at the University of Basel in Switzerland are looking to develop and even better system, called ‘vitrivr’.
Vitrivr, is the team’s in-development video retrieval system that has a unique method to find content. The user searches for a video from a database but, instead of typing out a name, the user can draw a basic sketch of the picture or video on a linked tablet to query it. Colouring in the basic shades of the background for example might throw up a few results, which can in turn be used to modify the search for better accuracy. If you’re still confused, just check out the demo video below.
Right now, vitrivr is being tested with a custom search engine and database, but it is scalable, meaning that in the future it can even be used with very large collections. Consider the possibility of the tool embedded into say Netflix. If you’re watching on your phone, you don’t even need to scroll out type out the show’s name, just draw the show’s logo and you find what you need.
The vitrivr system is open source and freely available on GitHub. And if you’re good at coding, and looking to try it out, there’s even a handy startup guide here (