Deep Learning Series - Session 4: Deep Dive - Advanced Neural Networks

Overview

In the first part, Dr. Yasuno will show how a CycleGAN is used for the automated monitoring of concrete damage in dams. Supervised learning requires a great deal of dataset and annotation works, so it takes long time to accumulate data source. In contrast, unsupervised deep learning approach for anomaly detection have progressed. Using the difference between the real damaged image and the generated fake output without damage, it is possible to detect the interest feature automatically. This session presents an anomaly detection application using unpaired image-to-image translation mapping from a damaged image to a fake output without damage using a generator network. Actually, we apply our method to field studies for monitoring dam surface using drone flight images.

In the second part, we will take a deeper dive into designing, customizing, and training advanced neural networks. We will demonstrate MATLAB's extended deep learning framework, which enables you to implement advanced network architectures such as generative adversarial networks (GANs), variational autoencoders (VAEs), or Siamese networks.

Highlights

• Comparing basic and advanced network architectures to determine the right architecture for your needs
• Generating realistic synthetic image data with GANs
• Implementing generalized research models in MATLAB

About the Presenters

Takato Yasuno received his D.E. degree from Tottori University. He has 18 years’ experience in 128 consulting projects as civil engineer for asset management planning. Since 2017, he works in the Research Institute for Infrastructure Paradigm Shift (RIIPS) as a senior researcher at the Yachiyo Engineering Company. His interest is Data Mining and Machine Learning, Image Processing for Structural Monitoring and Diagnosis, and Dam Inflow Forecasting for Flood Mitigation and Hydropower Opportunity. He is a member of JSAI. He has 10 machine and deep learning articles using MATLAB as listed in the URL. View LinkedIn Profile.

Paola Jaramillo is an application engineer at MathWorks in Eindhoven, Netherlands. She supports customers in finding the right solutions for doing signal processing with MATLAB and Simulink in different application areas including image processing, computer vision, and machine learning. Her primary interests are sensor data analytics, embedded systems, and self-adaptive systems. Before joining MathWorks, she was awarded a fellowship under an international double degree agreement with the Politecnico di Torino in Italy, where she graduated in Electronic Engineering. She carried out a six-month internship at IBM Zurich Research Laboratories in Switzerland, working on the design and implementation of a DSP module for FPGAs. During her research experience at Eindhoven University of Technology, she focused on sensor data analytics for intelligent lighting environments and actively participated in several European Commission projects.

Christoph Kammer is an application engineer at MathWorks. He supports customers in many different industries in the areas of machine and deep learning, image and signal processing and deployment to embedded or enterprise systems. Christoph has a master’s degree in Mechanical Engineering from ETHZ and a PhD in Electrical Engineering from EPFL, where he specialized in optimization and control design as well as the control and modelling of power systems.