We have developed a couple of web-based demonstrators for content-based video retrieval:


Lookapp, a system for automatic web based concept detector construction based on third party cloud computing services.


TubeTagger is a concept-based video retrieval system that learns to detect visual concepts (like "soccer", "desert", or "interview") from web video material.

Smart Video Buddy

Smart Video Buddy is our award-winning intelligent video recommender! Using latest computer vision technology, the system mines a video stream for semantic concepts and recommends content accordingly.


TubeFiler is an automatic genre categorizer for YouTube videos. The system assigns given tagged video clips to genre categories and semantic and visual clusters. We participated with TubeFiler in the ACM Multimedia Google Grand Challenge.


InViRe shows video-based similarity search for TV content. The system uses video fingerprints based on color, texture, and motion.


Navidgator realizes a visual browsing of image and video collections. Using a hierarchical clustering, content is arranged according to its visual similarity and can be browsed at different granularities (i.e. you can "zoom" into the image collection).
Image Understanding & Pattern Recognition Group
Multimedia Analysis and Data Mining Group