Computer Vision News - September 2020

of-the-art deep learning solutions and can apply them to their own datasets without the need to be a machine learning expert. This is an idea that is shared with a similar tool, Ilastik , which is already well established in the community. In his invited talk, Constantin Pape from the EMBL in Heidelberg shared the future plans around Ilastik: not only will deep learning be part of it, but Ilastik will also support models from the platform bioimage.io , amodel zoo for bioimage analysis, to bring progress made in the community to the hands of experimentalists. 3 BioImage Computing 61 In his invited talk, David Van Valen from the California Institute of Technology gave us insights how his group is addressing the eruption of large datasets and the consequent computational demands to process them. Using scalable pipelines with Kubernetes , David's group designed DeepCell , a platform for the segmentation and tracking of cells in real-world-sized datasets. Finally, Kristin Branson from HHMI Janelia and Virginie Uhlmann from the EMBL-EBI in Cambridge went full circle and showed us how to use computer vision and machine learning techniques in the context of downstream analysis. Kristin presented her group's work on the tracking of animal poses in videos and the subsequent characterization of behavior, while Virginie reminded us that not everything has to be deep learning: Although the analysis and classification of shapes has a long history in classical computer vision, their application to bioimages is of great value, non-trivial, and new efficient and applicable methods are needed for this domain. Best of ECCV 2020