Interview with Garrett Kenyon… Continued
Decoded Science: In many instances, neural networking research, such as Melanoma Diagnosis by Raman Spectroscopy and Neural Networks: Structure Alterations in Proteins and Lipids in Intact Cancer Tissue, has been applied to medical diagnosis. What medical applications do you foresee for improved visual recognition ability as the technology improves?
Kenyon: The most immediate application for our research would be in the design of better computer vision algorithms. For example, in detecting tumors in medical imaging data, perhaps as a pre-screening tool. More generally, we hope to contribute to the design of computer systems capable of synthetic visual cognition.
Decoded Science: If computer-based visual-recognition programs become highly-sophisticated and common, do you see it leading to today’s Captcha images becoming obsolete? Any ideas about potential replacements?
Kenyon: Very good question. It’s remarkable that there still does not exist software that can solve standard CAPTCHAS. Why? The amoeba/no-amoeba visual detection problem that we pose in our research project is, in fact, a type of CAPTCHA, although in this case, we have used lateral interactions to develop computer vision software that can solve this problem. More advanced CAPTCHAS will probably use natural images and the relationships between objects to represent more complex concepts. Solving such CAPTCHAS will require that computers be endowed with what can be operationally defined as synthetic visual cognition.
Artificial Vision: AI Capabilities
AI technology such as artificial neural networks are already in use in a number of industries. Medical applications, such as those mentioned above, may provide diagnostic assistance to reduce health care costs, and the use of facial-recognition software in everything from video games to crime prevention provide entertainment and improve safety measures. As technology improves, through projects like this one, the applications of shape and pattern recognition software offer opportunities to improve society in a variety of ways.
Gintautas, V., Ham, M., Kunsberg, B., Barr, S., Brumby, S., Rasmussen, C., George, J., Nemenman, I., Bettencourt, L., Kenyon, G. Model Cortical Association Fields Account for the Time Course and Dependence on Target Complexity of Human Contour Perception. (2011). Public Library of Science.
Brunelli, R. Template Matching Techniques in Computer Vision: Theory and Practice. (2009). Wiley.
Decoding Science. One article at a time.