ICMR 2015 Keynote: Multimedia Retrieval: Are We Doing it Right?

17 Jul 2015

June 24, 2015


Chua Tat-Seng


Multimedia has long been associated with visual media by many researchers. While visual media is one of the most challenging medium to deal with, the use of visual analysis alone is often insufficient nor necessary to tackle many real-life problems. The current wisdom seems to suggest that for real-world problem solving, it is necessary to make fuller use of content information including text and audio, and various kinds of contextual information, such as the user’s mobile device settings, social postings, check-in history etc., when available. This, for example, applies to social media and wellness research where a combination of text, visual, audio, health or other sensors, and public data are needed to fully analyze the problem. The second issue is the observation that big multimedia application ideas hardly come from the multimedia research communities. For example, many recent innovative applications in video like contextual advertising, live situational sensing, real-time news gathering, online navigation, wearable health devices, and personalized healthcare etc. have all come from the commercial world. While in research, we seem to barely even start on video research, as observed from the technical programs in recent ACM Multimedia and ACM ICMR conferences, where only a small fraction of visual papers is about video. This talk discusses issues in these two areas with examples, suggests directions to move forward as a esearch community, and outlines approaches to make our community more inclusive. The talk, however, will open up more questions than answers.

Click here for the slides.