Deep Learning Series - Session 2: Automated and Iterative Labeling for Images and Signals

Overview

In the first part, our guest speaker Raphaël Thierry from the Novartis Institute for Biomedical Research will show how he used Semantic Segmentation to pixel-label video frames in order to automatically process data from long-running experiments. This allowed Novartis to automate the time-consuming task of manually processing the videos, and helped the scientists to greatly increase the efficiency of their experiment analysis workflow.

In the second part, we introduce the signal, image and video labelers and discuss ways of extending these tools to facilitate labeling imagery to build AI models. We describe the use of preprocessing to facilitate the extraction of information from images, and present approaches to building models in an iterative fashion, validating predicted labels and incorporating on-the-fly models to label more and more data. We also discuss an approach to automating pixel-level labeling for semantic segmentation workflows.

Highlights

• Using and Extending the Signal, Image and Video Labelers
• Preprocessing to facilitate image labeling
• Iteratively building and incorporating computer-vision and machine-learning models
• Automating pixel-level labeling

About the Presenters

Raphaël Thierry is a Senior expert in Data Science at the NIBR (Novartis Institute for Biomedical Research) in Switzerland. He has authored over 50 publications in journals and international conferences and has been granted two patents in algorithmic design. He has an Engineer title in Computer Science and a Ph.D. degree in Applied Mathematics from the University Joseph Fourier. He held a postdoctoral research position for 4 years at ETH Zurich on the correction of photon scattering and beam hardening in X-ray CT. He then worked as a senior scientist at the center for cellular imaging and nanoanalytics (C-CINA), where he developed techniques for Electron Microscopy images. In 2012 he joined the Friedrich Miescher Institute in Basel, where he was responsible for image processing and software development at the Facility for Advanced Imaging and Microscopy (FAIM). Since 2018 he works at the NIBR on projects with a strong focus on image processing and computer vision involving deep/machine learning approaches.

Toon Weyens is an application engineer at MathWorks in Eindhoven, Netherlands. He supports innovative companies in automation and machinery, automotive, and aerospace industries by helping them use MathWorks software to analyze data, develop algorithms, create mathematical models, and scale to run on clusters, GPUs, and clouds. Prior to joining MathWorks, Toon was a postdoctoral researcher at the ITER Organization. He holds a M.Sc. degree in Energy Engineering from the University of Leuven, a M.Sc. degree in Nuclear Fusion Science and Technology from the Universidad Carlos III in Madrid, and a Ph.D. degree in Applied Physics from Eindhoven University of Technology.

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.