Workshop Program & Slides

Multimedia COMMONS 2016

Workshop Date: Sunday, October 16, 2016

Location: Roeterseiland Complex, University of Amsterdam

This Year's Theme: Datasets, Evaluation, and Reproducibility

 

9:00 - MMCommons Workshop Welcome

9:10 - Keynote Address: Roeland Ordelman, University of Twente & Netherlands Institute for Sound and Vision
    "Developing Benchmarks: The Importance of the Process and New Paradigms"

This talk will describe insights gained in the development process for several benchmarking programs over the last decade, as well as new angles seen in new benchmarking settings. View Abstract

10:10 - Plenary Session 1: The YFCC100M and the Multimedia Commons -- What Is It and How Can I Get it?

Brief introduction to the YFCC100M metadata corpus and expansions, to the Multimedia Commons corpora of computed features contributed by several institutions, and to subset annotations for location estimation (YLI-GEO) and event detection (YLI-MED). Presenters will summarize what’s in the datasets, how they were collected/extracted, and how to get them.

10:30 - Morning Tea/Coffee Break 

11:00 - Paper Session 1: Retrieval at Scale

Recent research using the YFCC100M and Multimedia Commons data explores how to both address and leverage the scale of the dataset for recognition, classification, and retrieval.
  • "In-Depth Exploration of Geotagging Performance Using Sampling Strategies on YFCC100M" -- Giorgos Kordopatis-Zilos, Symeon Papadopoulos, and Yiannis Kompatsiaris (CERTH-ITI) -- VIEW SLIDES
  • "YFCC100M HybridNet fc6 Deep Features for Content-Based Image Retrieval" -- Giuseppe Amato, Fabrizio Falchi, Claudio Gennaro, and Fausto Rabitti (ISTI-CNR) -- VIEW SLIDES
  • "Concept-Level Multimodal Ranking of Flickr Photo Tags via Recall Based Weighting" -- Rajiv Shah, Yi Yu, Suhua Tang, Shin'Ichi Satoh, Akshay Verma, and Roger Zimmermann (NUS, NII, UEC, and MNNIT) -- VIEW SLIDES

12:15 - Lunch Break

13:45 - Paper Session 2: Exploring the YFCC100M

Insights into the content and structure of the YFCC100M provide a basis for dataset comparison and for analysis of how people share media online.
  • "Analysis of Spatial, Temporal, and Content Characteristics of Videos in the YFCC100M Dataset" -- Jun-Ho Choi and Jong-Seok Lee (Yonsei U.) -- VIEW SLIDES
  • "Which Languages Do People Speak on Flickr? A Language and Geo-Location Study of the YFCC100M Dataset" -- Alireza Koochali, Sebastian Kalkowski, Andreas Dengel, Damian Borth, and Christian Schulze (DFKI and U. Kaiserslautern)

14:35 - Afternoon Tea/Coffee Break

15:05 - Interactive Panel and Discussion: Where Are the Opportunities for Research Impact?

What can we do with the YFCC100M that's new? What opportunities might we be missing? What tasks most benefit science and society? Panelists representing three broad areas of analysis (content, metadata, and social) will start the discussion off, then bring in the audience for a fishbowl-style discussion. Moderated by Martha Larson (TU Delft).

16:15 - Plenary Session 2: Future of the Multimedia Commons

An overview and open discussion of next steps for data, options for sustainability, and potential new tasks and challenges related to the YFCC100M and the Multimedia Commons.

16:40 - Adjourn