During this meeting, Prof. Ryan Rad (Vancouver Campus) will give us a talk on From Convolutional Neural Networks to Vision Transformers:  Advancing Earth Observation and Environmental Suitability. Details of the talk are given below.

You can join the Zoom meeting using this link:    https://northeastern.zoom.us/j/94841404890

Title: From Convolutional Neural Networks to Vision Transformers:  Advancing Earth Observation and Environmental Suitability

Speaker: Prof. Ryan Rad

Dr. Ryan Rad is an Assistant Professor at the Khoury College of Computer Sciences, Northeastern University in Vancouver. His teaching portfolio includes courses in Machine Learning, Algorithms, Software Engineering, and Data Analytics & Visualization at five different colleges and universities, reflecting his commitment to educating the next generation of computer scientists. With an extensive background in both academia and industry, Ryan’s career encompasses significant roles as a researcher, engineer, team lead, product owner, and advisor at a variety of startups and major corporations where he delivered innovative AI solutions across sectors ranging from healthcare to sustainability. In addition to his professional practice, Ryan actively contributes to the scientific community as a technical committee member and speaker at prestigious conferences and journals such as IEEE TIP, IEEE TMI, ICCV, and AAAI. Ryan is currently research-active at the intersection of machine learning and computer vision, and their application to environmental sustainability.

Abstract: This talk navigates the transition from Convolutional Neural Networks (CNNs) to Vision Transformers (ViTs) in computer vision, focusing on their application in Earth observation and environmental suitability analysis. We explore the evolution of CNNs, highlighting key milestones and innovations, before delving into the architecture and advantages of ViTs. Through seminal works like ViT, we uncover how ViTs excel in capturing long-range dependencies and contextual information. Finally, we discuss their application in tasks such as land cover classification and wildfire spread modeling, showcasing their transformative potential in advancing environmental research and sustainability efforts.