Audio and Multimedia Researchers Win Competition with Crowdsourcing Proposal

^{July 2, 2013}
Audio and Multimedia researchers Luke Gottlieb and Benjamin Martinez Elizalde have been selected as winners of the Ideas Competition sponsored by the Workshop on Crowdsourcing for Multimedia, which will be held in conjunction with ACM International Conference on Multimedia in October. The competition asked for short crowdsourcing proposals related to multimedia problems. Crowdsourcing, in which a large number of people perform simple tasks in exchange for a small incentive, provides an affordable way to complete repetitive tasks and also allows the managers of the task to get results from a diverse group of workers. Gottlieb and Martinez's proposal is one of ten selected for funding. The researchers will ask crowdsourced workers to classify videos by simple binary conditions - for example, whether a video was shot indoors or outdoors or whether it has music. The results of the work will help other multimedia projects by providing an audio profile of the videos shot under each condition.

Recent slide on ICSI's event detection work

In event detection work, for example, ICSI researchers are building a system capable of searching a large body of videos for specific events, such as "feeding an animal." Some events are more likely to occur in specific conditions, such as outdoors. Data collected from the crowdsourced workers will help establish a profile for videos shot under certain conditions. See, for example, the ALADDIN Project. Audio profiles for indoor and outdoor videos could help in work on geo-location, in which a system tries to automatically place where a video was shot. It might also help researchers to know, for example, when videos have music, which makes audio cues more difficult to detect.