Juyang Weng - US grants

This is the first year of a three-year continuing Research Initiation Award. The resarch aims to develop theiries and techniques of automatic learning for object recognition from 2D images. Visual recognition is recognized as a veruy difficult task due to large image variations, high complexities, and a huge amount of data in the images that computers muyst deal with in the real world. Automatic learning is crucual for achievement of versatile and reliable vision systems. The new approach to learning in this research project is general, flexible, adaptive, and efficient. The proposed methods will be tested on a large number of real-world images to recognize learned objects and identify their locations in the images.

The university is establishing a Cooperative Multimedia Computing Laboratory to support undergraduate education in the Department of Computer Science. The laboratory is using state-of-the-art equipment to create a new multimedia curriculum based on three courses: Software Engineering, Computer Graphics, and Computer Networks. Laboratory assignments are being designed to allow students to gain experience with large-scale, real-world applications involving multimedia software, multimedia interfaces, and multimedia communication. Initial laboratory sets are being drawn from environmental science, computer-assisted medicine, and automated manufacturing. The faculty involved in the establishment of the proposed laboratory have extensive experience in these application areas and in developing laboratory exercises that expose students to new technologies and computing methods. The Department of Computer Science, the College of Engineering, and the university are very committed to this project. Funds have already been allocated for a preliminary, small-scale version of the laboratory, summer salaries and course reductions for the three principal investigators, laboratory space, supporting personnel, and long-term maintenance of the equipment.

9617321 Stockman, George C. Jain, Anil K. Michigan State University CISE Research Instrumentation: Equipment for Real-Time Interaction with Images This research instrumentation grant enables the purchase of equipment to be used in the following research projects: -Self Organization for Vision-Based Learning, Databases and Content-Based Retrieval, Modeling , Tracking, and Interpretation of Human Faces. Research is being performed in vision-guided robotics, visual learning, organizing and accessing databases of images and modeling and tracking human faces. Key elements of the research include - investigation of systematic methods for automatic learning over multiple tasks including navigation and object recognition, -organization of imagery for fast access using integrated features such as shape, texture, and color, and -modeling and storage of objects for fast recognition and real-time tracking. Advancement in the research projects depends critically on the new instrumentation which will provide the capability to interact in real-time with both sensed images from the environment and images stored in memory. A steerable camera system will be used for experiments in navigation, hand-eye coordination and tracking of hands and faces for gesture recognition. RAID disk memory connected to two powerful workstations will provide for large and fast storage for two simultaneous experiments which may be storing or searching video images or performing real-time model-based tracking. We anticipate that our research projects will yield several benefits, including improved man-machine interaction, better access to image databases, and better methods for learning automatic recognition procedures needed in many practical problems.

The objective of this research is an initial exploration of developmental learning by machines. This is motivated by human cognitive development from infancy to adulthood. It requires a fundamentally different way of addressing the issue of machine intelligence. The basic goal of the proposed developmental learning algorithm is to enable a machine to develop its cognitive capability, after its "birth," through direct interactions with its environment. The developmental learning concept does not mean just from small to big and from simple to complex. It requires the system to learn new tasks and, as a special case, new aspects of a complex task without a need for reprogramming. The basic nature of developmental learning plays a central role in enabling a human being to incrementally scale up his or her level of intelligence from ground up. In the proposed exploratory study, the major basic concepts for developmental learning will be crystallized and some preliminary experimental results will be produced for the focused task category of recognition through interactive training.

IIS-9900498
Weng, John
Michigan State University
$30,000.00 - 15 mos.

Workshop on Development and Learning

There is a growing interest in the study of cognitive and behavioral
development and the interactions between what is innate and what is
learned during the development. New theories and architectures for
development are being studied in fields related to both artificial and
natural intelligence. Scaling up from ground, both in size and functionality,
is required for dealing with real-world problems and for better
understanding natural intelligence. How does an individual, biological or
artificial, scale up its cognitive and behavioral capabilities through interactions
with the environment? What are the common mechanisms
that enable scaling up for a variety of cognitive and behavioral
capabilities and their integration? Since this important subject is
interdisciplinary, this workshop will bring together researchers from
closely related fields, including artificial intelligence, machine learning,
computer vision, pattern recognition, speech recognition, robotics,
animal learning, developmental psychology, neuroscience and computational
linguistics. The aim is to discuss (1) the role that development and
learning can play in the development of artificial and natural intelligent
agents, (2) the common principles that are shared by the development for
very diverse cognitive and behavioral capabilities such as vision, speech,
language, reasoning, planning, decision making, locomotion and
object manipulation. (3) important directions for future research on
development and learning, and (4) short-term and long-term applications.
Findings of the workshop will be documented in a report which will be widely
disseminated in the various related research communities. A website on
development and learning will be created from this workshop as a long-term
cross-disciplinary information center for those who are interested in
artificial and natural intelligence.