June 6, 2018, 12:15pm-1:15pm
Title: Visual Recognition of Symbolic and Natural Patterns
Speaker: Dr. Alexandra Branzan Albu, UVic Department of Electrical and Computer Engineering & Computer Science
Join DHSI and UVic Libraries to hear about computer-assisted pattern recognition during this special lunchtime talk
Abstract: Image-based object recognition is a visual pattern recognition problem; one may characterize visual patterns as either symbolic or natural. Symbolic patterns evolved for human communication; they include but are not limited to text, forms, tables, graphics, engineering drawings etc. Symbolic patterns vary widely in terms of size, style, language, alphabet and fonts; however, literate humans can easily compensate for this variability and instantly recognize most symbolic patterns. On the other hand, natural patterns characterize images of physical structures; they often lack the intrinsic discriminability and structure of symbolic patterns, and vary widely in terms of pose, perspective, and lighting.
This talk will explore similarities and differences in approaches designed for recognizing visual and symbolic patterns, and will address the following questions via examples:
- What are the distinctive characteristics of natural patterns? What dimensions of variability can we infer?
- What are the distinctive characteristics of symbolic patterns? What dimensions of variability can we infer?
Dr. Alexandra Branzan Albu is an Associate Professor with the Department of Electrical and Computer Engineering and cross-listed with Computer Science. Her research interests are related to image analysis, computer vision, and visual computing. She is actively pursuing outreach activities dedicated to increasing the women’s presence in electrical engineering and computer science.