Exploring low level image processing to boost middle and hig
iHARP Talk Tuesday Series presents Dr. Xavier Soria
Guest Speaker
Dr. Xavier Soria, Research Scientist, Adjunct Professor at the National Discovery of Chimborazo, and an Associate Researcher at the ESPOL Polytechnique University
Bio
Dr. Xavier Soria is a research scientist working in image processing and computer vision. His current research focuses on advancing the field of the image Edge Detection and thermal image Super Resolution driven by Artificial Intelligence (AI) models. He is interested in developing lightweight and efficient models. His contributions have transcended the initial domains and are now being applied to various high-level computer vision applications. Currently, he is an Assistant Professor in the Polytechnic School of Chimborazo at Morona-Santiago and an Associate Researcher at the ESPOL Polytechnic University, both in Ecuador.
Presentation Title
Exploring low level image processing to boost middle and high-level computer vision tasks
Presentation Abstract Since the advent of automatic image processing in the 60s of the 20th centuries, enormous advances have been made. Several fields have made use of computer vision and image processing tasks as a result of these breakthrough topics, from smile detection in smartphone's cameras to space exploration in the vast universe. The first part of this presentation will introduce low-level image processing from the very beginning. I will briefly summarize the image edge detection task, highlighting current investigation and advances in this field. Secondly, I will present the contribution of my team in the edge detection domain; particularly, I will focus on how we see the problem and what advantages our proposal has over existing state-of-the-art models. Finally, I will show how edge detection boosts and improves work in different fields, such as breast cancer detection, calculation of ice sheet thickness from radar images and so on. All the models and datasets presented in the talk are freely available