VideoPlus: a platform for augmented video content
Computational Science and Engineering, Data Sciences
Computational Shape Analysis
Video content uploaded on social video sharing systems such as YouTube, Vimeo, or Dailymotion has experienced an exponential growth in the past years, making them the favorite medium for advertisement. The ease of sharing and accessing video content has also provoked significant efforts on the side of content creators and owners aimed at protecting their video assets from illegal reproduction and distribution, including the distribution through different channels, the use of digital right management (DRM) encrypted systems, and video fingerprinting systems for the detection of illegally uploaded video content.
The ambition of VideoPlus is to develop a close-to-market system that would serve as a “uniform resource identifier” for video content, employing technologies which are based on video fingerprinting to precisely synchronize spatio-temporal metadata with any video content, without being limited to any specific video source or distribution channel. VideoPlus builds on three main components: (i) the algorithmic core, result of the ERC-funded project COMET, where we developed an efficient fingerprinting algorithm capable of extracting a short binary signature (hash) of image and video content and instantaneously locate it in a very large dataset; (ii) the back-end, relying on specifically tailored data structures to store and retrieve spatio-temporal information (e.g. video subtitles and annotations) together with frame hashes and align it with the video; (iii) the front-end, able to capture the video content, extract its hash, communicate with the back-end to retrieve the spatio-temporal metadata, display them synchronously with the video, and allow user interaction.
ERC Proof-of-Concept grant 737548;