Die Kosten für Tutorials sind bis zu 31.01.2019 reduziert und ab dem 01.02.2019 zu regulärem Preis weiter verfügbar.
Genaueres können Sie der Registrierungsseite entnehmen.
Der Anmeldeschluss für Tutorials ist der 15.02.2019.


Prof. Dr.-Ing. habil. Andreas Maier
Pattern Recognition Lab, Friedrich-Alexander-Universität, Erlangen, Germany

Deep Learning Fundamentals

Deep learning has received a lot of attention in the machine learning community. Successful applications from speech recognition or computer vision are already part of our daily life. Much effort has been devoted to transferring this success to medical image computing. Therefore, neural networks have become an essential research direction.
The first half of this tutorial is designed to familiarize participants with neural networks. The second half presents the transition from neural networks to deep learning.
The building blocks of classical neural networks, such as the multi-layer perceptron, activations and loss functions, are explained. Furthermore, the concepts of gradient-based learning and backpropagation to calculate the gradients are introduced.
The second part of the tutorial covers the elements of convolutional neural networks, around which most successful deep learning applications revolve. Special attention is devoted to regularization techniques, which are essential to state-of-the-art performance. Best practices and exemplary architectures conclude the tutorial.

PD Dr. Klaus Maier-Hein
Division of Medical Image Computing, German Cancer Research Center (DKFZ), Heidelberg, Germany

Advanced Deep Learning Methods

The remarkable rise of Deep Learning has led to an overwhelming amount of new papers coming up by the week. This tutorial intents to filter out the research most relevant for the medical image computing (MIC) community and present it in a structured and understandable form. It is composed of five parts: Classification, Segmentation, Detection, Generative Models and Semi- Supervised Learning. Each part starts off with a thorough motivation, shows exemplary use cases related to MIC, provides a brief model overview and describes the current state-of-the-art methods in the respective area.
Basic knowledge about Neural Networks as covered by the “Deep Learning Fundamentals” tutorial is recommended.

Jun.-Prof. Dr. Mattias Heinrich,
Unversität zu Lübeck

Hands-on Deep Learning in pytorch.
Deep learning is currently often a prerequisite to achieve state-of-the-art performance for supervised classification tasks in medical imaging. Theoretical knowledge of DL fundamentals and the discussion and use of advanced architectures from current research are very important and are already covered in two excellent BVM tutorials.
Nevertheless, modern software architectures enable an accompanying process for practical understanding and thus are very appealing to young as well as established researchers and doctoral students to go beyond the current state-of-the-art and effectively explore and implement own ideas.

In this tutorial, we would like to give a hands-on exercise for implementing your own deep learning network with particular focus on remote training on cloud hardware, manual layer definition for unconventional operations, and the use of GPU and DL-optimisers for classical image analysis tasks, such as deformable registration. Basic knowledge of neural networks and python is sufficient for participants – pytorch – the modular framework of our choice is user-friendly and easy to learn within few hours.