Elisa Barney Smith is a professor at Luleå University of Technology (LTU), Sweden which she joined after a 22 year career in the Electrical & Computer Engineering department at Boise State University in Boise Idaho USA. She is in LTU’s Machine Learning group in the Department of Computer Science, Electrical and Space Engineering, Embedded Internet Systems Lab (EISLAB). She received a B.S. in Computer Science and the M.S. and Ph.D. degrees in Electrical, Computer and Systems Engineering all from Rensselaer Polytechnic Institute, Troy, NY, USA. Professor Barney’s main research interests are image processing and machine learning. She applies these primarily to document imaging as well as to facilitate image processing for disparate areas from biomedical image processing to materials science research to soil remediation evaluation. Her research in document analysis has included developing models of the degradations produced during document image acquisition, analyzing the defects that can be produced by the models, ballot image processing, handwritten Indic script recognition, and historical document image processing, recognition and analysis. Some of this work was featured in a TEDxBoise talk (available on YouTube). Professor Barney has worked on numerous cross-disciplinary, multi-institutional, national and international project teams. Her expertise has given her invitations as a paid guest scientist at the NATO Saclant Center in Italy, the Ecole Nationale Supérieure des Télécommunications in Paris, France, LORIA in Nancy France and the Technical University Dortmund in Germany. Professor Barney co-authored a textbook on machine learning. She has received NSF, NASA, and industry funding, including a prestigious NSF CAREER award. Elisa Barney Smith is a member of International Association of Pattern Recognition (IAPR) and a Senior Member of IEEE and SPIE. She is past chair of the IEEE Boise section, past IEEE Region 6’s North East Area chair, the 2019 & 2020 IEEE (global) Student Activities Committee (SAC) Chair, and has served on several global IEEE committees.
Elisa.Barney@ltu.se
https://sites.google.com/view/drbarney/
https://www.linkedin.com/in/elisa-barney-smith-4a222432
DVP term expires December 2025
Presentations
Applying Electrical Engineering and Computer Science to the Humanities
Applications of electrical engineering and computer science like software engineering, computer hardware, circuits and power grids are well known. The tools developed by engineers and scientists can also be applied to the humanities in fields such as art, literature and history. This has become known as Digital Humanities. In her career Elisa has worked to apply image processing and machine learning (a more accurate word for the overused term “Artificial Intelligence”) to many problems in many fields, and recently several in Digital Humanities. Those projects include analyzing handwriting in WWI postcards, analyzing marginalia written by 19th century American author Herman Melville, studying typography in early printed books, helping historians extract information from medieval manuscripts, and even extracting text from a Dead Sea Scroll. This talk containing many vivid illustrations will introduce the audience to the field of Digital Humanities and describe work she has done in some of these projects.
Document Analysis – Reading between the Lines
Documents are all around us. Currently most documents are born digital and many may remain in digital form for their entire lifespan. Other documents appear on paper at some point in their lives, and there is often a need to bring them into the digital world. Most camera systems come with an Optical Character Recognition (OCR) module. On cleanly printed and clearly photographed documents, documents with simple layouts convert to digital form well. For other documents, a significant amount of research remains. There is more to the process than just recognizing individual letters. The process starts with image processing, then the letters and words are recognized. The relationships between letters, words, sentences, paragraphs and pages need to be understood. Equations and figures have a two-dimensional structure that needs to be interpreted and then related to the document structure. This talk will discuss the various components in a Document Image Analysis (DIA) system, including some of the machine learning techniques used in the process, and how processes vary for modern versus historical documents, and machine print versus handwritten documents.
Do you believe in Camelot?
Everyone wants to create documents with perfect quality, but image degradations are pervasive. The sources are many. Many degradation models have been proposed with the aim of improving the image quality. To challenge image recognition systems and to provide adequate quantities of training data synthetic datasets are purposely created with embedded degradations. How do the perfect and the distorted cohabitate? In this talk an overview of some image degradation models from simple to detailed will be presented, followed by discussion on how they are often used in image processing and in image recognition. The need for synthetically created degradations to match degradations seen in real data and how models can be validated will also be covered. * Cultural note: Camelot is a fictional realm in English folklore where everything was perfect, until it wasn’t. The talk will accommodate audience members without the cultural background and explain the concept as needed to connect the talk title with the talk content.
