DIGITAL

Multimodal Deep Learning was a key issue at the "MMM 2019".

DIGITAL presented the latest research results in the field of Multimedia Analytics at the "25th International Conference on MultiMedia Modeling in Thessaloniki".

Multimodal Deep Learning was a key issue at the "MMM 2019".
Video Browser Showdown at the MMM 2019. Credit: JOANNEUM RESEARCH/Bailer

From 8th to 11th of January 2019, one of the leading forums for researchers and industry practitioners to share new ideas, original research results and practical development experiences from all multimedia-related areas took place. The MMM program included 119 contributions presented as oral presentations, posters, tutorials or demos, and the Video Browser Showdown session, resulting in a rich program extending over four conference days. Main conference topics included multimedia indexing, multimedia fusion methods, media representation and algorithms, audio, image, video processing, coding and compression, mobile multimedia applications, multimedia web applications, and much more.

Dr Martin Winter and Werner Bailer, Key Researchers at the Institute DIGITAL contributed to the conference with two oral presentations. Martin Winter presented an approach for a training face recognition algorithm online, enabling the recognition of persons with few (even a single) labelled example. The method can also be used to automatically train the face recognition system for unknown persons encountered in the image/video content, and associating the detections later with a name without rerunning the analysis of the content. Werner Bailer presented a method for replacing faces in image content with artificially generated ones (using generative adversarial networks, GANs) in order to enable use and storage of the images while preventing privacy issues resulting from the presence of identifiable faces. This presentation was part of a special session on multimedia analytics, which aims at using automatic multimedia analysis tools in order to support human decision-making. The session also included a panel discussion with all presenters, focusing on topics like privacy, content verification and usability of multimedia retrieval.

The Video Browser Showdown (VBS), co-organized by Werner Bailer, is an annual live video search competition that is part of MMM since it was hosted in Klagenfurt in 2012. At VBS researchers evaluate and demonstrate the efficiency of their exploratory video retrieval tools on a shared data set in front of the audience. This year six teams from Europe and Asia competed to find video clips defined by visual or textual queries in a 1,000-hour video collection within a few minutes. The competition included “novice” rounds, where members of the audience operate the tools, in order to assess their usability for non-experts. Both presentations and the Video Browser Showdown had very positive feedback from the audience, motivating to visit the  MMM 2020 conference and VBS in Daejeon, Korea.

 

Links
 

MMM 2019

Video Browser Showdown – The Video Retrieval Competition

Connected Computing @ DIGITAL