WIAMIS 2008: technical notes (Part 1)

As promised, this is the first part of a series of postings on the technical side of WIAMIS 2008 (before anyone starts thinking that the conference was all about socializing and visiting beautiful places. :-) 

It refers to the morning portion of the first day.  

The day started with Horst Bischof's excellent invited talk on "Robust Person Detection for Surveillance using Online", where he managed to engage everyone -- even those who are not directly working in this field -- with a great speech, fully supported by excellent slides, video clips, and demos. His take-home message was "Person detection in surveillance scenarios is feasible using online learning methods", but what struck me the most was his team's approach to conservative learning, which could be summarized as follows: "it is better to get fewer, but good, examples of persons".

The next session (Oral Session I: Analysis I) had four papers, three of which are very relevant and related to work that my students, collaborators, and I have been doing during the past few years:

• "A Comparative Study of Classification Techniques for Knowledge-Assisted Image Analysis", by G. T. Papadopoulos, K. Chandramouli, V. Mezaris, I. Kompatsiaris, E. Izquierdo, and M. G. Strintzis
• Follows up on the group's work reported recently in a rich and inspiring Eurasip JASP paper.
• "A Semantic Multimedia Analysis Approach Utilizing a Region Thesaurus and LSA", by Evaggelos Spyrou, Giorgos Tolias, Phivos Mylonas, and Yannis Avrithis
• Explores a combination of the (growingly popular) RSST segmentation algorithm, selected MPEG-7 descriptors, K-means clustering, and LSA in video documents.
• I'm curious to know where the next stages of their work (which plan to include context) will take them.
• "Exploiting Spatial Context in Image Region Labelling Using Fuzzy Constraint Reasoning", by Carsten Saathoff and Steffen Staab
• Learning and modeling spatial context is a hot topic. In this work (by a world-famous Semantic Web-oriented group) it is approached from a Fuzzy Constraint Reasoning perspective.

These papers (and associated references) have pointed me in directions that I had not considered before and given me many ideas that I hope to put into practice soon. 

If you're reading this blog and are interested in these topics (semantic image annotation and classification), please drop me an email or a comment on this entry.