Скачать или смотреть Mobile Search and Visualization of Historic Multimedia Data

The omnipresence of smartphones and their increasing capability of capturing high quality photos and videos leads to ever-growing multimedia collections. While those multi-sensor devices not only capture the scenery, they also annotate the picture with lots of meta-data, particularly where and when the data was produced. Yet the possibilities in browsing and searching these collections along their spatio-temporal meta-data is very limited. Alongside the recently created auto-annotated multimedia collections, archives around the world are digitalising historic photos. However their meta-data are mostly not as extensively available as photos taken by smartphones. Additionally, the integrated camera of the smartphone allows capturing a live stream of the real world in front of it. In this thesis, we present an approach on content-based multimedia retrieval focused on spatio-temporal image retrieval. We further list the retrieval results in comparison to the user's current location. Combining the previously mentioned live stream with the retrieval results lead to a superimposition of the current view and an image. Additionally, we notify the user once an overlay is possible, based on the current location. We call the resulting mobile application Go Find!. In this thesis, we describe our approach on the development of the mentioned application. We evaluate the resulting prototype using historic data from the city of Basel and analyse the outcome of the evaluation.

BSc Project by Loris Sauter