Call for papers
Aims and scope
The focus of the workshop is on the unique opportunities, challenges, and solutions for robust multimedia learning across different domains. Since the revolution in web information retrieval, multimedia analysis and retrieval has been regarded as the next grand challenge. Media-sharing sites like Flickr, YouTube, and Last.fm have brought huge amounts of multimedia resources, reflecting all aspects of social life, with open access to anyone. Such explosions of multimedia data and associated tags provided by amateur-consumers, bring new opportunities for researchers to deepen our already acquired understanding. So far, our knowledge has been restricted to specific domains including sport video, news video, and natural images. The proposed workshop strives to broaden our understanding in this emerging area, with an ultimate aim to make unstructured multimedia data from broad domains accessible, reusable, searchable, and manageable.
Multimedia learning
As a corner stone in this field, machine learning techniques are densely employed and models are built for tasks varying from categorizing images into scenes, detecting semantic concepts in video, and recommending music by their content similarity, to name a few. Given the variety of domains where multimedia is generated these days, there is a clear demand to generalize and adapt models trained from one domain to other domains, both for scientific, technological, economic, and performance considerations. For example, the amateur-tagged images with partially, and ambiguous labels may serve as a useful resource for novel video analysis and retrieval applications. Another scenario might be to utilize semantic concept models trained for news video, for detection in video surveillance. Current research has shown impressive results in multimedia analysis for semantic concepts, but evaluations have repeatedly shown that these good results do not transfer well to similar concepts in other domains. Thus, robust learning approaches are in urgent demand given the potential impact they may produce and the abundance of existing but unexplored labeled multimedia resources available.
Topics of interest
We solicit submissions that introduce key concepts, discuss theoretical frameworks and technical approaches, challenges, research opportunities, and open issues in areas of interest in robust multimedia learning, including (but not limited to) the following areas:
- structured domain modeling algorithms;
- robust content and context dependent learning;
- online/social multimedia analysis and mining;
- transfer learning with human-in-the loop;
- transfer learning models in multimedia analysis/retrieval;
- evaluation and benchmark metrics for broad domain multimedia learning;
- success and failure analysis for robust multimedia learning;
- multimedia learning systems and applications in and cross domains of biometrics, mobile media, personal media, and the arts, among others;
- future challenges for robust multimedia learning;
Paper submission
All authors are requested to use the workshop EDAS system for sending in their submissions. The pdf submissions should be up to 8 pages in the ACM style in English. All submissions will be peer-reviewed by at least 3 members of the program committee.
ACM proceedings style
All the submissions should be prepared in double-column ACM proceedings style. The corresponding latex and word files can be found from the following links:
ACM classification
Please note that each paper must provide a classification of it's content and general terms according to the ACM computing classification. If you are not familiar with this classification please make sure that you read carefully through the guide "How to classify works using the ACM's computing classification system" to select the most suitable categories and general terms for your paper and add some additional keywords to it. Do NOT omit classification, general terms or keywords in your camera ready copy, e.g., in order to save space!