Machine Learning / Deep Learning Workshop

Machine Learning APPLIED to biological image analysis

Duration: 2.5 days, i.e. 17.5 hours
Objective: Understand and apply machine learning methods for the analysis of biological images using "free" software.
Prerequisites: Know the basics of image analysis
Audience: Researchers, engineers, technicians, doctoral and post-doctoral students

Program:
Theory and practical training:
This workshop is intended for biologists who do image analysis. During the two days of training, the basics of machine llearning methods will be presented and participants will be able to apply them to their own problems. The workshop will alternate between theoretical courses and practical work. The practical session will use free software such as ImageJ/FIJI (weka-segmentation), Ilastik, Icy and Cellprofiler Analyst, each tool offering different solutions for machine learning.

The first day will be dedicated to classical learning methods using for example k-means clustering, random trees or support vector machines. These methods can be used for dense cell segmentation, object tracking, 3D structure segmentation and cell counting.

The second day will be devoted to deep learning. Using a simple example, we illustrate the creation, training, validation and use of a network. We will show how to use transfer learning to adapt existing networks to your own applications. Concrete examples of the application of "Deep Learning" for image analysis are: segmentation of bacteria, motion tracking without markers, 3D segmentation, PALM / STORM assisted by deep-learning, catering and classification of tissue images in histology.
For this second day we will work with the “tensor flow” library of Python. Knowledge of programming would be a plus.

Please check the upcoming Biocampus workshops for the next date the workshop will be held and for inscription.

You can find the slides and exercises of the workshop here.