The concept of the research conducted in our laboratory is a task-oriented vision: First, we determine a task to make the computer recognize the real world; next, we find the problem to be solved in order to carry out the task; then, we think about how we solve the problem; and finally, we develop a novel algorithm and theory.
Digitally Archiving Cultural Heritages
Our laboratory is conducting the project, which digitally preserving by measuring the cultural heritages all over the world with many kinds of state-of-art laser sensors in the precision of around 2-3mm. From now on, we have archived the Kamakura Great Buddha in Japan, the Nara Great Buddha in Japan, the Wat Sri Cham Great Buddha in Thailand, and Bayon Temple in Cambodia. These digital data have a wide area of applications, such as for CG contents, for the information for repairing of the cultural heritages if it is lost or deteriorated, for the study of the history of the cultural heritages by virtually restoring the original appearance of them, etc.
Learning the Technique of Artisans by Robot
Infants first imitate the parents' actions as the first stage of the learning. The purpose of our laboratory is to automatically generate a robot program which observes the human action by vision system, recognizes the action, and reproduces the same action. In particular: (1) correcting the observation error of the object trajectory, (2) extracting the essence of the task from observation, (3) recognizing and reproducing the hand manipulation, and (4) manipulating a soft object (rope) by the robot. We are planning to permanently preserve the technique of artisans by learning the technique of living national treasures by the robot with these proposed methods.
Modeling the Traditional Dance and Performing the Dance by Humanoid Robot
We are conducting a research to model the motion of the dance in order to succeed many kinds of Japanese traditional dances to posterity in order to solve the problem that there are only few successors of these dances. We first recognize the human dance motion by computer, and then reproduce the dance performance by 3DCG or humanoid robot.
Physics Based Computer Vision
Currently, PBCV (physics based computer vision) is considered to be important compared to the past computer vision research which concentrates on the geometrical property. We are conducting a research of PBCV, such as developing the shape estimation method of transparent objects by analyzing the polarization phenomena, and separating the highlight component from the object image which has complex textures.
Intelligent Transportation System (ITS)
We are developing a method for processing image/range data by using electric maps, cameras, and range sensors, in order to generate the 3D model of the wide area of the real world such as towns and cities. We are also conducting a research such as to attach the movement in the real world to the virtual world, and to classify and recognize the car and the type of the car.
Dr. Ikeuchi recieved IEEE R&A K-S Fu Memorial best transaction paper award in 1998, and his paper, "Numerical Shape from Shading and Occluding Boundaries," was selected as one of the most influential papers to have appeared in the Artificial Intelligence Journal within the past ten years. He recieved David Marr prize in ICCV 1990, and also recieved several awards at CVPR 1991, VSMM 2000, VSMM 2004. He recieved various best paper awards from Japanese professional societies such as Robotics Society of Japan and Japan Virtual Reality Society, and also recieved the best paper awards at the Japanese conference named MIRU 2000.
General chairs: IROS 1995, ITSC 1999, IV 2001. Program chairs: CVPR 1996, ICCV 2003.
Editor-in-Chief: IJCV. Editorial board: IEEE TPAMI, IEEE R&A, CVGIP, JOSA.
IEEE fellow. Distinguished Lecturer of the IEEE Signal Processing Society, 2000-2001. Distinguished Lecturer of the IEEE Computer Society, 2004-2006.
|Research Interest Keywords|
|computational sensors, computer graphics, computer vision, humanoid robotics, range data, range finders|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions