The draft notebook paper for TRECVID 2009 by the MediaMill team, containing members of the University of Amsterdam, INESC-ID and the University of Surrey, is now available. In the paper we describe our TRECVID 2009 video retrieval experiments. The MediaMill team participated in three tasks: concept detection, automatic search, and interactive search. Starting point for the MediaMill concept detection approach is our top-performing bag-of-words system of last year, which uses multiple color descriptors, codebooks with soft-assignment, and kernel-based supervised learning. We improve upon this baseline system by exploring two novel research directions. Firstly, we study a multi-modal extension by the inclusion of 20 audio concepts and fusing using two novel multi-kernel supervised learning methods. Secondly, with the help of recently proposed algorithmic refinements of bag-of-words, a bag-of-words GPU implementation, and compute clusters, we scale-up the amount of visual information analyzed by an order of magnitude, to a total of 1,000,000 i-frames. Our experiments evaluate the merit of these new components, ultimately leading to 64 robust concept detectors for video retrieval. For retrieval, a robust but limited set of concept detectors necessitates the need to rely on as many auxiliary information channels as possible. For automatic search we therefore explore how we can learn to rank various information channels simultaneously to maximize video search results for a given topic. To improve the video retrieval results further, our interactive search experiments investigate the roles of visualizing preview results for a certain browse-dimension and relevance feedback mechanisms that learn to solve complex search topics by analysis from user browsing behavior. The 2009 edition of the TRECVID benchmark has again been a fruitful participation for the MediaMill team, resulting in the top ranking for both concept detection and interactive search.

multocida, and its activity against B.
An interesting ‘Viewpoint’ appeared in this month’s issue of Communications of the ACM. In Time for computer science to grow up, Lance Fornow argues that because the field of computer science at large is maturing, we need to switch our attention from publishing deadline-driven conference papers, often conforming to the “least-publishable unit”, to well worked out archival journal papers. I cannot agree more. Lance provides several arguments in favor of publishing journal papers rather than conference papers, but ignores one important one. When computer science researchers are evaluated among scientists from more traditional disciplines , reviewers typically compare their published journals rather than their conference papers. As conference papers are often considered unimportant in other disciplines. The paper is online at:

Toky Tower by night

This week I finally had the opportunity to visit Japan. I attended the Pacific-Rim Symposium on Image and Video Technology 2009 for lecturing our tutorial on concept-based video retrieval. With more then 50 people present during the lecture, the tutorial was well attended. I was happy to see that the topic of video search is also of scientific interest on this part of the globe. Apart from interaction with people giving talks, posters and demo’s at the conference, I had some spare time to explore Tokyo and even made a short sightseeing day trip to Nikko. My experience with this part of Japan is quite positive. Tokyo is well organized, clean, and efficient, and despite the language barrier finding your way in this metro pole is relatively straightforward. In cases you do get lost the folks around here are always willing to help. Ooh, and did I mention the quality of the food already? I was lucky that my hosts took me out to some fancy Japanese restaurants for traditional dinner, including sake of course. Today I finished my trip with a visit to Shin’ichi Satoh‘s lab at NII, presenting the latest achievements of our team in TRECVID and social image retrieval. After, again, a marvelous dinner it is now time to fly back home, but I hope to return to Japan in the near future.

Two of my colleagues have just released, independently, software for computation of color descriptors. Jan-Mark Geusebroek released Color Sift, based on a recently accepted CVIU paper, and Koen van de Sande released the ColorDescriptor software, related to our 2008 CVPR and CIVR papers. These color descriptor have proven to be highly effective under many circumstances. See also our performance in the TRECVID 2008 video retrieval benchmark, and our winning position in the PASCAL VOC 2008 object classification competition. This software might be useful for many people working on visual concept detection, amongst others.

N’importe qui peut acheter des médicaments maniaco-mélancoliques sans prix dans la pharmacie régionale de votre propriété sous Lamictal ouvertement. Et obtenez le meilleur livre pour manic depression drugs. Cette pharmacie fournit des pilules de grande qualité. Dyspeptiques phénomènes, des selles et une diminution de l’activité enzymatique du foie sont possibles dans le tractus gastro-intestinal Un manque d’efficacité de la dose de Lamycatalum peut provoquer de rares cas de rhabdomyolyse, coagulation intravasculaire de cellules de sang, le syndrome de plusieurs défaillances d’organes Lamictal Les Interactions Médicamenteuses Compétitif du métabolisme de la lamotrigine avec les enzymes hépatiques, ralentit son assimilation.
The ACM MIR 2008 paper entitled “Learning Tag Relevance by Neighbor Voting for Social Image Retrieval” by Xirong Li, Cees Snoek, and Marcel Worring is available online. Social image retrieval is important for exploiting the increasing amounts of amateur-tagged multimedia such as Flickr images. Since amateur tagging is known to be uncontrolled, ambiguous, and personalized, a fundamental problem is how to reliably interpret the relevance of a tag with respect to the visual content it is describing. Intuitively, if different persons label similar images using the same tags, these tags are likely to reflect objective aspects of the visual content. Starting from this intuition, we propose a novel algorithm that scalably and reliably learns tag relevance by accumulating votes from visually similar neighbors. Further, treated as tag frequency, learned tag relevance is seamlessly embedded into current tag-based social image retrieval paradigms. Preliminary experiments on one million Flickr images demonstrate the potential of the proposed algorithm. Overall comparisons for both single-word queries and multiple-word queries show substantial improvement over the baseline by learning and using tag relevance. Specifically, compared with the baseline using the original tags, on average, retrieval using improved tags increases mean average precision by 24%, from 0.54 to 0.67. Moreover, simulated experiments indicate that performance can be improved further by scaling up the amount of images used in the proposed neighbor voting algorithm.