These are the photos from the 27th IST seminar on March 8, 2012.
The number of participants: approximately 8 (including 2 presenters).

(1) presentation by Dr. Kazuaki Nakamura

(2) presentation by Dr. Zhaozheng Hu

(3) discussions

->the complete archive of the pictures 1
->the complete archive of the pictures 2

Time/Date: 13:00-15:00, 08 Mar
Location: Room 304, Research Building #5 (総合研究５号館).
  http://www.media.kyoto-u.ac.jp/ja/access

First slot:

Talk: “Recognizing Ingredients and User’s Actions in Cooking Situations for a User-Centric Cooking Support System” (in Japanese)
Speaker: Teaching Assistant Kazuaki Nakamura

Abstract:
Our research team aims to realize a “user-centric” cooking support
system, with which a user can cook in a normal way freed from the constraints of the system. Nowadays people’s interests in foods and
cooking are growing in Japan and several prototypes of the cooking
support systems have been developed. However, most of the current
systems have adopted a “system-centric” manner, in which a user is
constrained by the system to cook in a pre-determined way. This will
frustrate the user, whereas our “user-centric” system does not lead to
such a frustration. The “user-centric” system should understand what the
user is doing in each moment during cooking, which requires at least the
techniques for recognizing ingredients and the user’s actions. Our
current status about the ingredient recognition and the action
recognition will be reported in this presentation.

Second slot:

Talk: “Relay Visual Surveillance with Pair Active Cameras for Gaze Computation” (in English).
Speaker: JSPS fellow, Zhaozheng Hu, (Matsuyama Lab.).

Abstract:
Computation of human gaze usually requires good-resolution and　 well-calibrated images, which is an extreme difficult task in a wide surveillance space. We address this problem by proposing the relay visual surveillance (RVS) system that consists of pair active cameras. In the system design, the problems of camera layout and surveillance space partition with non-uniform cells are solved by maximizing the facial image resolution while meeting the various constraints including mechanical delay of active cameras, visual coverage, object resolution, etc. A cell-based method is applied for RVS system calibration. The active camera control and switch scheme is also presented to detect and track a moving object in relay. The proposed RVS system has been tested in an indoor surveillance experiment, where a fully gaze perception in a wide surveillance space was successful.

These are the photos from the 26th IST seminar on December 8, 2011.
The number of participants: approximately 17 (including 2 presenters).

(1) presentation by Dr. Shuichi Miyazaki

(2) presentation by Ms. Mayumi Kamada

(3) discussion

->the complete archive of the pictures

Time/Date: 13:00-15:00, 08 Dec
Location: Room 304, Research Building #5 (総合研究５号館).
http://www.media.kyoto-u.ac.jp/ja/access

First slot:

Talk: “Approximation algorithms for the stable matching problem” (in Japanese)
Speaker: Associate Prof. Shuichi Miyazaki, (Okabe Lab)

Abstract:
The stable matching problem is the problem of finding a “stable” matching for given preference lists of participants. This problem is used in many situations, such as assigning residents to hospitals, university students to laboratories, and so on. In this talk, I give some basic introduction of the stable matching problem, as well as the results on the approximation algorithms for it.

Second slot:

Talk: “Prediction of protein residue contacts and structure-function analysis for time-series data of protein structure.” (in Japanese).
Speaker: PhD candidate, Mayumi Kamada, (Akutsu Lab.).

Abstract:
Proteins have an important role in life activity. They can serve their functions only by forming their structures and interacting with other molecules. Since interaction between protein residues is a clue for understanding these mechanisms, many methods have been developed to predict protein residue contact from amino acid sequence. In this talk, I will introduce our prediction method for predicting protein residue contacts based on co-evolution information using random field model. Proteins are dynamic molecules within the cell, and the dynamics is thought to play a crucial role for structure change and interaction between molecules. For correlation analysis between structure motion and protein function, we extract motion features from time-series data of protein structure using wavelet transform. I will show some analyses using the features.

These are the photos from the 25th IST seminar on November 10, 2011.
The number of participants: approximately 11 (including 2 presenters).

(1) presentation by Ms. Natalia Efremova

(2) presentation by Mr. Yamato Okamoto

(3) discussion

->the complete archive of the pictures