vitrivr is an innovative multimedia retrieval engine that offers a rich set of different query paradigms to search in large collections. This includes traditional keyword queries but also query-by-example, query-by-sketch, and motion queries that rely on intrinsic vide features features such as, for instance, color, edge or motion information. At the user interface, sample images or sample videos (query-by-example), hand-drawn sketches (query-by-sketch) or flow fields (motion queries) and combinations thereof can be used. The system is composed of the Vitrivr-NG front-end, the flexible retrieval engine Cineast, and the storage back-end ADAMpro.
Vitrivr-NG is vitrivrs browser-based user interface. It is implemented using Angular.js and Typescript and its modular architecture enables it to be customized for many use cases.
Cineast is a content-based multimedia retrieval engine which retrieves images, audio- and video sequences as well as 3d-models based on their content. It supports Query-by-Example as well as Query-by-Sketch by using a multitude of low-level visual features in parallel.
ADAMpro is a database system to store and retrieve multimedia data. It provides Boolean retrieval and similarity search and makes use of a various index structures for efficient retrieval.
- Twitter: To apply for one of our projects, we ask you to
- state the problem you try to solve
- state the required steps
- provide a detailed working plan
Furthermore, we ask you to submit your code to one of the pre-projects mentioned on our GSoC page.
See also http://www.vitrivr.org/gsoc.html#student-application for more details on what we expect.