(Semantic-based HIerarchical Automatic Tagging of videos by Segmentation using cUts)
SHIATSU is a general framework for the smart management of large video collections.
The goal of SHIATSU is to provide efficient and effective tools for:
- automatically segmenting a video into visually coherent scenes (possibly also exploiting audio content), representing each scene by way of a set of representative keyframes;
- automatic tagging of keyframes/scenes/videos with simple textual tags and/or semantic tags, i.e., concepts extracted from taxonomies/ontologies;
- searching scenes/videos by way of textual/semantic tags and/or visual similarity (e.g., “give me the scenes tagged with "Leonardo DiCaprio" which are visually similar to "this" picture of a ship”).
|
SHIATSU has been applied in several real contexts, among which a digital interactive application involving movies. A collaboration with Cineteca di Bologna has also been started for the enhanced management of digital video archives.
|
|
since March 2014 |
Send comments to i.bartolini@unibo.it Last updated: |