��������

Course Description

The goal of this course is to familiarize researchers in the life sciences with state-of-the-art deep learning techniques for microscopy image analysis and to introduce them to tools and frameworks that facilitate independent application of the learned material after the course.

The following topics will be covered extensively during lectures, exercises, and project work:

1. Image Denoising and Restoration (Fully supervised and self-supervised),
2. Image Translation (i.e., Generating fluorescent-like images from label-free acquisitions),
3. Image Segmentation (Various flavors will be presented and explored),
4. Image Classification,
5. Object Detection and Tracking in 2d and 3d Time-Lapse Movies, and
6. Explainability Methods
7. Popular Models (e.g., U-Nets, GANs, Transformers)

The course will be organized into two one week phases. Week 1 is a lecture- and exercise-based phase that starts by introducing basic and later also more advanced concepts of deep learning and allows participants to learn in detail how existing state-of-the-art methods and tools operate under the hood. The second week is project-based, where students will work together with numerous experts to apply the newly acquired knowledge and skills to their own datasets and analysis problems. Faculty and TAs will assist the students in data preparation, problem formalization, network architecture design, tool selection, model training, prediction, and evaluation.

Students will leave the course with an appreciation for the power and limitations of deep learning as well as broad knowledge of key tools that apply deep-learning based methods to microscopy image data.

The course assumes familiarity with the Python programming language, but does not assume any prior experience with machine learning techniques. To ensure that students can “hit the ground running”, we will provide pre-course materials and exercises that will enable students to assess their programming skills and introduce basic data handling and Python coding basics. Students are highly encouraged to bring their own microscopy datasets to work on during the second week, ideally with some amount of existing label data that can be used to train adequate deep learning models. This course is limited to a maximum of 24 students.