MediaEval 2013 - International Workshop on MediaEval Benchmarking Initiative for Multimedia

In 2013, MediaEval is offering six classic tasks and six Brave New Tasks. The tasks are listed below with links to the individual task pages where you will find more detailed description of the task definition, target group, data set and evaluation metric. Note that the timelines can vary from task to task, so please refer to the individual task pages for data release and submission dates.
The MediaEval 2013 worksop will take place in Barcelona, Catalunya, Spain, on Friday-Saturday 18-19 October 2013. Note that this is just before ACM Multimedia 2013 http://acmmm13.org/, which will be held Monday-Friday 21-25 October 2013, also in Barcelona. Workshop registration will open in July.
A limited number of student travel grants in the amount of 500 Euros will be available to support travel of outstanding student participants to the workshop. More information about this opportunity will be available on this site in July.
Tasks offered in 2013:
Social Event Detection for Social Multimedia This task requires participants to discover social events and organize the related media items in event-specific clusters, within a collection of Web multimedia. Social events are events that are planned by people, attended by people and for which the social multimedia are also captured by people. Read more...
Search and Hyperlinking of Television Content This task requires participants to find relevant video segments to an information need and to provide a list of useful hyperlinks for each of these segments. This year we focus on television data provided by the BBC and real information needs from home users. Read more...
Placing: Geo-coordinate Prediction for Social Multimedia The Placing Task requires participants to estimate the geographical coordinates (latitude and longitude) of media items (images and videos), as well as predicting how “placeable” a media item actually is. The Placing Task integrates all aspects of multimedia: textual meta-data, audio, image, video, location, time, users and context. Read more...
Violent Scenes Detection in Film (Affect Task) This task requires participants to automatically detect portions of movies depicting violence. Participants are encouraged to deploy multimodal approaches (audio, visual, text) to solve the task. Read more...
Visual Privacy: Preserving Privacy in Surveillance Videos For this task, participants will need to propose methods for protection of privacy sensitive elements (i.e., obscuring identifying elements on people) in videos so that they are rendered unrecognizable in a manner that is suitable for computer vision tools and is perceived as appropriate to human viewers of the footage. Read more...
Spoken Web Search: Spoken Term Detection for Low Resource Languages The task involves searching FOR audio content WITHIN audio content USING an audio content query. This task is particularly interesting for speech researchers in the area of spoken term detection or low-resource speech processing. Read more...
(New!) Question Answering for the Spoken Web The problem that we wish to explore in this task is how best to build an information retrieval system in which both the queries and the content are spoken. The task challenges the research community’s ability to build ranked retrieval systems for matching spoken questions with spoken answers based on topical matching. Read more...
(New!) Soundtrack Selection for Commercials (MusiClef Task) Given a TV commercial, participants are required to predict the most suitable music soundtrack from a list of candidate songs. A multimodal dataset will be provided involving both context- and content-based information, such as audio features, visual features, web pages, social tags and microblog information, related to brands, products, artists and songs. Read more...
(New!) Similar Segments of Social Speech This task involves searching in social multimedia, specifically conversations between students in a university department. This task is the first exploration of social search in multimedia, and the first social spoken dialog retrieval task not assuming term-based search. Read more...
(New!) Retrieving Diverse Social Images This task requires participants to automatically refine a ranked list of Internet photos using provided visual and textual information. The objective is to select only a set of photos that are equally representative matches but also diverse representations of the query. Read more...
(New!) Emotion in Music (also an Affect Task) This task is a new task on emotional characterization of music. Given a set of songs, participants are asked to automatically generate emotional representations. Read more...
(New!) Crowdsourcing for Social Multimedia This task requires participants to create ground truth from raw labels that have been generated by workers on a commercial crowdsourcing platform. Read more...
Tasks marked "New!" are the 2013 Brave New Tasks. If you sign up for these tasks, please be aware that you will be asked to keep in close touch with the task organizers concerning the details of the task over the course of the benchmarking cycle. We ask for extra-tight communication in order to ensure that these tasks have the flexibility they need to reach their goals.