Shigemi Nagata

Mobile robot control by a structured hierarchical neural network

A mobile robot whose behavior is controlled by a structured hierarchical neural network and its learning algorithm is presented. The robot has four wheels and moves about freely with two motors. Twelve sensors are used to monitor internal conditions and environmental changes. These sensor signals are presented to the input layer of the network, and the output is used as motor control signals. The network model is divided into two subnetworks connected to each other by short-term memory units used to process time-dependent data. A robot can be taught behaviors by changing the patterns presented to it. For example, a group of robots were taught to play a cops-and-robbers game. Through training, the robots learned behaviors such as capture and escape.<<ETX>>

An images-based 3d model retrieval approach

This paper presents an images based 3D model retrieval method in which each model is described by six 2D images. The images are generated by three steps: 1) the model is normalized based on the distribution of the surface normal directions; 2) then, the normalized model is uniformly sampled to generate a number of random points; 3) finally, the random points are projected along six directions to create six images, each of which is described by Zernike moment feature. In the comparison of two models, six images of each model are naturally divided into three pairs, and the similarity between two models is calculated by summing up the distances of all corresponding pairs. The effectiveness of our method is verified by comparative experiments. Meanwhile, high matching speed is achieved, e.g., it takes about 3e-5 seconds to compare two models using a computer with Pentium IV 3.00GHz CPU.

Similarity-based partial image retrieval system for engineering drawings

Designers of mechanical products frequently refer to engineering drawings which are stored as image data in databases to design a new mechanical product efficiently. Multiple mechanical parts are usually drawn on each engineering drawing. Therefore designers want to find engineering drawings containing parts similar to a query image in the shape of a part drawn on an engineering drawing. In this paper, we propose a novel similarity based partial image retrieval system for engineering drawings. A unique aspect of this system is that a graph representation is utilized to robustly find engineering drawings containing similar parts which are invariant to the size, position, and rotation. We verified the performance for the similarity based partial image retrieval system through experiments using industrial engineering drawings. The results show that the top five similar engineering drawings for every query image are always accurately retrieved by our proposed system. This finding suggests that this system could be useful for the reuse of stored engineering drawings.

MIRACLES: Multimedia Information RetrievAl, CLassification, and Exploration System

This paper describes our multimedia information retrieval system, MIRACLES, for retrieving multimedia content such as web pages, business documents, and movies. MIRACLES creates information units, which are collections of related media items, such as an image and text explaining the image. MIRACLES arranges and displays the information units in 3-D space so that similar information units are located near each other. The user can narrow down the retrieval of the multimedia content by rearranging the information units in the display space. MIRACLES is already in practical use, and is employed in some commercial products and services. We are presently evaluating the use of MIRACLES by actual users.

Control of a multivariable system by a neural network (inverted pendulum)

An inverted pendulum presents a typical problem of controlling a nonlinear multivariable system. In the paper a discussion is presented of inverted pendulum control by a neural network. The neutral network learns a control law through the trial and error method. In the simulation and actual system, it was possible to invert the pendulum stably at the target position after some trial and error.<<ETX>>

SVM-based active feedback in image retrieval using clustering and unlabeled data

In content based image retrieval, relevance feedback has been extensively studied to bridge the gap between low level image features and high level semantic concepts. However, it is still challenged by small sample size problem, since users are usually not so patient to label a large number of training instances. In this paper, two strategies are proposed to tackle this problem: (1) a novel active selection criterion. It takes into consideration both the informative and the representative measures. With this criterion, the diversities of the selected images are increased while their informative powers are kept, thus more information gain can be obtained from the feedback images; and (2) incorporation of unlabeled images within the co-training framework. Unlabeled data partially alleviates the training data scarcity problem, thus can improve the efficiency of SVM active learning. Systematic experimental results verify the superiority of our method over some existing active learning methods.

Owlery: a flexible content management system for “growing metadata” of cultural heritage objects and its educational use in the CEAX project

With the Educational use of Cultural heritage Archives and Cross(X) search (CEAX), we have investigated how to establish a framework for managing various kinds of information on cultural heritage objects and how to utilize them for educational purposes. To achieve this goal, we propose a conceptual framework in this paper called “Growing Metadata” and a flexible content management system called Owlery. Growing Metadata includes not only factual descriptions of objects but also various annotations about the objects, such as metadata for children, course materials prepared by school teachers, classroom reports, etc., and are reusable for search and educational purposes. Owlery is a software platform to create, share, utilize and reuse the Growing Metadata, and in which various metadata and annotations are managed in different levels of authenticity, authorship, and user groups. As a result of the experimental classes for 89 6th-grade children, our framework was found to be efficient and accepted by the content creators, like museum experts, content annotators and shool teachers.

Mobile robot control by neural networks using self-supervised learning

A reinforcement learning algorithm based on supervised learning is described. It uses associative search to discover and learn actions that make the system perform a desired task. One problem with associative search is that the systems actions are often inconsistent. In the searching process, the systems actions are always decided stochastically, so the system cannot perform learned actions more than once, even if they have been determined to be suitable actions for the desired task. To solve this problem, a neural network that can predict an evaluation of an action and control the influence of the stochastic element is used. Results from computer simulations using the algorithms to control a mobile robot are described. >

A sensory information processing system using neural networks

In order to carry out actions particular to the goals, a robot processes sensory information, that is, it transforms sensed data to internal representation. In some cases, the robots internal representation cannot be determined uniquely from the sensed data. An architecture is proposed for a sensory information processing system that overcomes this ill-posed problem. The system uses an artificial neural network which is trained to transform internal representation to sensory data. Applying an iterative scheme to the network, the unique internal representation can be determined. The scheme compares the networks output (sensory data) with the sensed data, and by backpropagating the difference through the layers updates an input (internal representation) which could have created the applied output (sensed data) based on the gradient descent method. By predicting the resulting state based on the intention of the systems own movement, the accuracy and speed of sensory information processing can be improved. Simulation results for three-dimensional object recognition are given.<<ETX>>

Shape based 3D model retrieval without query

We describe our shape based 3D model retrieval method that is based on a browsing technique. With this method, users can retrieve the desired 3D model efficiently without a query model. In previous retrieval systems, users should provide a 3D model as a query to the system. Then, the system retrieves similar 3D models and returns them to the user. However, the problem of how to obtain the query model remains. With our method, users can retrieve the desired 3D model by walking through a virtual 3D space without the query model. At first, 3D shape features are extracted from all the 3D models, and 3D models are arranged and classified in the virtual 3D space so that similar 3D models are placed near each other. This allows the user to easily grasp where the 3D models similar to the desired one are located. After approaching the 3D model that is similar to the desired one, users can focus on all the nearby models, which are usually similar to the desired one. So users can find the desired one efficiently. We also developed two functions to make our method more efficient. Firstly, our method needs to render a large number of 3D models at one time quickly, so we developed a high-speed rendering method. Secondly, to make it easier for the user to choose the desired one from many 3D models, we developed a method to make 3D models face the direction from which users can recognize the shape of the 3D models easily. In addition, we present the results of experiments to evaluate the retrieval efficiency, which shows that our method is four times as fast as a retrieval method using a query model.