Overall Goals

The overal goals are:

1. To build a large scale thesaurus well-spread over the semantic clues
2. To design, adapt and evaluate methods to learn large thesauri of detectors
   1. Active learning for large thesauri
   2. One-class classifiers for large thesauri
3. To define and evaluate powerful sets of visual, audio, and cross-modal invariant features
   1. Visual colour and local shape features
   2. Texture and contexture features
   3. Object-level features
   4. Motion pattern features
   5. Acoustic features
   6. Multimodal integrated features
   7. Selection
4. The objective is to deliver effective interaction with the user
5. The objective is to evaluate the approach in relevant application areas. Areas of application: video news, documentaries and surveillance video