Extremely short and really delayed ACM MM 2000 / IBM Watson trip report Date: Wednesday, 29 Nov 2000 By: Frank Nack ACM 2000 - October 30th - November 04 Since we had no paper (and I think it was not too bad having none) I used the time to talk to as many people as possible by still keeping an eye on the presentations. In general: the conference was quite boring and in my humble opinion and didn't show anything new. Most of the presentations provided the audience with improved versions of algorithms or a different combination of methods for retrieval, not that the results would be better. Another problem: most talks provided exactly the content of the paper, including formulas etc. => so most of the time it was better reading than listening. Conversations with participants revealed, that most were astonished not to see more meta-modeling and higher level representation presented at the conference. Furthermore, it was mentioned quite often that one cannot really see any longer what the presented work is good for (Michael Vernick from Bell brought it to the point: most work presented is too complicated and can unfortunately only be applied once the material is produced - which does not help in real time applications). Thus the general feel was that the comunety has to achieve a connection between high level structures and low level features - where I thought great, that is what we are doing, to some extend. It was also suggested to apply for a tutorial and/or panel for the ACM 2001 in Canada. Thus, we should consider arranging that. Presentations I found kind of interesting were: Yong Rui's presentation about the automatic extraction of highlights for TV baseball programs (though I wonder how many non-Americans will make use of it). It would be interesting to see how their audio-based approach can be combined with what was done at CSRIO (where I believe that their work is better, because it covers a larger set of sports). William Bares' presentation on Virtual 3D Camera Composition from Frame Constraints, which demonstrates a tool for planning camera shots. Afrodite Sevasti's annotation environment for Young children was quite cute - not that the talk was great, but I liked the tool. Key notes: The computer revolution hasn't happened yet by Alan C. Kay from Disney The most interesting statement: nearly everybody in this conference is concerned with the question 'how' things should be done but that is wrong, you should rather ask 'why'. Showed an interesting tool - Squeak: see squeak.org, an easy to use environment for children to explore physics in a playful way (though I sometimes thought that it is a bit difficult for children). Nevertheless it is quite interesting from the interaction point of view. Enabling Next generation Streaming Media Networks by Eric A. Brewer, Inktomi Corporation Main aspects: importance of caching, multicast and bandwidth provisioning and most important network edge services. Quite some time of the talk was devoted to business models, where his idea was that the content providers will pay to the server providers, so that their content will be supplied by the edge servers (an edge server is a customer oriented server, thus in between server and client, situated at the edge of the network). His business model was attacked most by the audience - where the current understanding was that the service provider will pay the content providers, because they will have the interesting material, which finally is more important than the delivery technology. Panels: Curricula and Resources about Multimedia - Ed Fox, Wolfgang Effelberg, Nicolas Gerorganas, Rachelle S. Heller and Ralf Steinmetz The panel was very disappointing. The general assumption of the panelists was that the curriculum for mm for computer scientists (and that was all they were interested)must include compression, feature detection, networking, etc and might provide information on hci, meta data modeling, perception theory, philosophy. Only Rachelle objected and suggested that students should in general learn about the function of each medium, the relation between the different media and the use for different media in diverse applications. The audience provided similar ideas, i.e. that the domain the mm application is focused on is of importance - meaning that mm development is always application and domain dependent, that students have to know about copyright problems and should get an idea about the economics. The panel finished off with the statement of the panelists that they actually do not really know how to develop good mm software for learning and they would be happy if the audience would provide them with input. They suggested the following sides: CRIM http://www.cistc.org and JERIC http://fox.cs.vt.edu/JERIC/index.html Multimedia Copyright Enforcement on the Internet - M. Potkonjak, James M. Burger, Christopher J. Cookson, Darko Kirovski, David P. Maher, Jeremy Welt Nothing we have not heard before. However, it was interesting to see that the current discussion on NAPSTER does not really apply on an industry such as film - they really do face different problems, due to their different production/business model. The general feeling was that watermarking does not really solve the problems and that we need something different than writing some very small IPR info directly into the data. Nevertheless, a very entertaining panel with lots of accusations and sarcastic opponent attacks. Demonstrations: Well, all so nice toys with the most boring graphics or well tuned videos. Those I liked were (descriptions can be found in the proceedings): Mediacaption - a demo by Florian Mueller Interactive Tools for Constructing and Browsing Structures for Movie Films by Riad Hammound and Roger Moore Video Scouting Demonstration: Smart Content Selection and Recording by Nevenka Dimitrova et al. Mediacaptain by Florian Müller see http://www.mediacaptain.com/ Workshop: Bridging the Gap: Bringing Together New Media Artists and Multimedia Technologists (http://www.lucentcache.com/workshop.html) The ws was quite interesting, in particular that artists do think in different ways about mm as the work being presented during the conference might suggest. Our paper was well received and some of the artists showed interest in collaboration (Betty Beaumont - New York University, Teri Rueb, Dept. of Visual Arts, U Maryland Baltimore County, and Jim Mahoney). The proceedings (conference and ws) are in our office. There are also two tutorial proceedings available: Instructional Design for MM and Scalable Multimedia Servers. ----------- Talk at IBM Watson I spent two days at IBM Watson and had a look at their mm server architecture. Quite nice - and it is working. What they offer is the recording of a talk in real time with two cameras, where one is reacting on changes on the screen and thus capturing the slide-show, whereas the other camera tracks the presenter. The software automatically merges the two images into one video stream, which is automatically stored on the server. There is an additional stream for the audio track. The material is then be additionally annotated later on. Their streaming server is used indoors for allowing IBM employees worldwide to look at presentations given at the different IBM sites. Furthermore, the server is used for the IBM internal education program. The talk I gave was again well perceived and is, due to the fact of their server technology, not only available for IBMer but also as an mpeg file on a CD-Rom in our office (I received it an hour after I had finished the presentation). Frank