ilastik: Faster and More User-Friendly Through Full Pyramid Support
Anna Kreshuk (European Molecular Biology Laboratory)
To enable multi-scale interactive machine learning on large datasets in ilastik through full exploitation of state-of-the-art pyramidal file formats and viewers, and extend functionality to other bioimage analysis tools.
ilastik is a popular open source tool which brings machine learning-based image analysis to users without computational expertise. The intuitive UI of ilastik and its carefully optimized algorithms have made it the go-to tool for interactive image segmentation in the bioimage analysis community. The on-going development of the deep learning workflow and the bioimage model zoo helps us attract new users in microscopy and beyond. The goal of this proposal is to enable interactive user-friendly machine learning on multiscale data for all users of ilastik and, through a Python API, for computational researchers and developers of other tools. In this work, the team will focus on the technical and engineering challenges of introducing the scale dimension and the corresponding UI to ilastik. Ultimately, ilastik users will be able browse large multi-scale datasets, annotate structures of interest, propagate annotations across scales, train different classifiers at different scales and combine their outputs in a cascaded manner.