From Sep. 23rd to Sep. 28th I visited several sessions of the 15th ACM Multimedia Conference in Augsburg, Germany. I have a feeling that researchers are trying desperately to make computers automatically understand the semantic of images and videos. A lot of papers in the conference were on image segmentation and image annotation, there was a whole-afternoon-session, all the presentations were more or less on image annotation powered by machine learning, lots and lots of Bayesian network, Markov fields stuff. The best paper award went to the paper “Correlative Multi-Label Video Annotation” written by Guo-Jun Qi (USTC China),
Xian-Sheng Hua (Microsoft Research Asia), Yong Rui (Microsoft China R&D Group), Jinhui Tang(USTC China), Tao Mei (Microsoft Research Asia) and Hong-Jiang Zhang (Microsoft Adv. Tech. Center) (See the picture). Artificial intelligence plays the central roll here. Here is the page for the session “Tutorial on Bayesian Methods for Multimedia Signal Processing”.
There was a very interesting presentation session for open source projects, the winner of the best open source software was Hop, a new functional programming language designed for Web 2.0. The paper can be found here. The Hop guys coined the web broker concept. A web broker sits between multimedia content providers and users, and provides elegant solution to highly interactive web application (such as web gallery, web media player, etc) programming. Hop is not designed for n-tier web application programming, however.
PS: Sorry for the delay of evaluation of the Drools Puzzle Round 2. I think Dr. Starke, the winner of last round, is too busy these days. I have not heard from him for long. But eventually the submissions will be evaluated by Dr. Starke, and the third round will come out. I will have more time for community activities from November on, so if till then there is still no result for the second round, I will evaluate the submissions and get the Drools Puzzle on track again.