This half-day presentation is designed for students, researchers and engineers at all experience levels. Whether you are new to the field or an experienced practitioner, the material is structured to bridge the gap between theory and real-world implementation, making it accessible and valuable to a broad audience.
What You Will Learn
By attending this presentation, participants will:
- Understand the fundamentals of sonar image formation and projection geometry.
- Explore deep learning approaches for 3D reconstruction from sonar imagery.
- Learn how generative models, such as CycleGANs, can help overcome limited training data in real-world underwater environments.
Topics Covered
- Foundational Sonar Theory: Participants will learn the basics of 2D sonar image formation and projection geometry. No extensive background in acoustics is required.
- 3D Model-Based Reconstruction: The presentation will introduce reconstruction methods for simple scenarios based on image-to-image affine transformations, which can serve as building blocks for more advanced techniques.
- Application-Focused Deep Learning: Deep learning topics, including convolutional neural networks (CNNs), transformers and generative models, will focus on practical applications and the challenges of underwater perception in low-visibility environments.
Intended Audience
Graduate students, academic researchers and deep learning engineers seeking practical insights into modern AI techniques for underwater sensing, imaging and perception.
Attendees are encouraged to bring a laptop to participate in simulator walkthroughs and software demonstrations.
