Multi-Modal Multimedia Retrieval Blocks at Their Core (Master Thesis, Ongoing)


Rahel Arnold


The goal of this thesis is the development of newly designed multi-modal multimedia retrieval building blocks and their integration and evaluation in vitrivr.

The thesis concentrates on the foundations of multi-modal multimedia retrieval. For this objective, a holistic retrieval model is created in the rst step, which can be applied to di erent media data, such as images, videos, or text. The retrieval search in the de ned model is then related to k-Nearest-Neighbour-, text-, Boolean- and metadata search. The system is based on the ndings of vitrivr, the retrieval procedure currently used by the DBIS group. A retrieval engine and a user interface are implemented in a second step founded on the constructed model. Further on, the new system and the model will be evaluated. For this purpose, a user evaluation is planned, structured in the style of a combination of VBS-like and LSC-like tasks.

Start / End Dates

2022/04/11 - 2022/10/10


Research Topics