Application of Machine Learning to Computer Vision and Medical Imaging
Recent developments in Machine Learning have fostered applications in computer vision and
medical imaging. We demonstrate ongoing work of our lab in this field:
(1) The continuously increasing amount of available imagery (photographs on the web,
overhead satellite images, surveillance videos) calls for efficient algorithms for automated
categorization and content-based retrieval. We present a state-of-the-art approach to
categorize a large variety of real world images (B. Ommer).
(2) Nowadays reconstructing 3D-structures from electron microscopy images requires
cumbersome manual annotation by highly-trained experts. To aid understanding the brain’s
connectivity, we develop machine learning techniques for the automated reconstruction of
neuronal processes from transmission electron microscopy sections (V. Kaynig).
(3) Tissue microarrays (TMA) combined with immunohistochemical staining, provide a high-
throughput method of analyzing potential biomarkers in a large cohort of patients.
Classification of TMAs requires highly-trained pathologists and has profound impact on
clinical decisions. We develop methods to aid and objectify this process (T. Fuchs).
Participants have the opportunity to interact with the researchers developing the aforementioned
methods and algorithms. Participants experience the challenges contemporary image processing
faces, how machine learning addresses these issues, and how automated image analysis is of
increasing relevance for many aspects of research and daily living.