To provide image/video access as an information service and entertainment on web and mobile platforms. Large digital archives are currently accessed by meta data. Integrating efficient content-based access will enlarge the usability of the archives used by different communities.
This WP will focus on semantic search and services to support event, news and entertainment analysis on the Internet and mobile phones especially for video platforms such as zie.nl owned by Ilse Media. Although Ilse Media currently delivers a platform for video storage, indexing and search are limited to key words. Content-based search is urgently needed. Ilse Media lacks the knowledge or expertise on content- based video access. Large scale datasets will be made accessible by Ilse Media and a beta-site will be issued to gain experience in “semantic video search in the wild”. This provides this WP with documented datasets, user-driven questioning, and content interaction. Semantic search has so far been targeted at a single information source; this WP combines multiple modalities to go beyond standard video retrieval. The modules developed by UvA are integrated in a demo system by Ilse Media and made available for public use (mobile and desktop), increasing the societal and economic value of semantic video search.
mobile video search issuing more requirements on community and sharing of video.
Description of work
- Task 9.1: Based on user study, valuable concepts are identified for different application areas (UvA, Ilse Media).
- Task 9.2: techniques and software are developed for large scale processing of features and classifiers for concept and event recognition in videos suited for web (zie.nl) and mobile phone platforms (UvA, Ilse Media).
- Task 9.3: techniques are developed for image quality analysis and shot representation for video processing and summarization (UvA).
- Task 9.4: information fusion is developed for video search based simultaneously on text, annotations, speech and pictorial modalities (UvA, Ilse Media)