MM

Building flexible, accurate multimedia analysis and retrieval systems requires massive amounts of annotated data as ground-truth input. The Multimedia COMMONS workshop will lay the groundwork for developing a research community around the Multimedia Commons Project (MMCP), an initiative initially focused on annotation of---and research using---the 99.2 million images and nearly 800,000 videos in the Yahoo Flickr Creative Commons 100 Million (YFCC100M) dataset released in 2014. This freely available dataset is already being used as the basis for several derived corpora and related tools that form the Multimedia Commons, including the YLI Corpus, a public-domain corpus of computed audio, visual, and motion features; YLIMED, a set of public-domain annotations specialized for multimedia event detection; and the YFCC100M Browser, a tag-based query tool for visualizing the distribution of YFCC100M data.

We invite all current and potential users of the YFCC100M and associated Multimedia Commons resources (read: everybody!) to attend MMCommons. Researchers currently using the MMCP datasets---or comparing other large datasets---will present their work in paper sessions and demo systems built for the ACM MM Grand Challenge. The papers presented will address the challenges of leveraging massive amounts of multimedia data from a variety of angles, from evaluating the use of tags for training to extracting social meaning from related images.

Much of the workshop will focus on maximizing the long-term value to the multimedia research community of these new resources. In plenary sessions and panel and discussion formats, workshop attendees are invited to actively participate in shaping the future of the Multimedia Commons Project. Topics will include how to leverage the dataset to stimulate new research directions through challenges and benchmarks and how to extend the impact through applications in other fields (e.g., anthropology or medicine). These explorations will be grounded in a discussion of how we get there: what are the needs and priorities of multimedia researchers with regard to annotation and feature data, and how can we build a collaborative structure for developing and distributing that data.

In laying the groundwork for focused collaborations around an open, shared dataset, we hope to inspire participation from a diverse set of multimedia researchers working on a diversity of tasks. We particularly hope to encourage research that ties together multiple research threads or approaches questions from multiple perspectives.