InstantDL - An easy-to-use deep learning pipeline for image segmentation and classification

摘要: MotivationDeep learning contributes to uncovering and understanding molecular cellular processes with highly performant image computing algorithms. Convolutional neural networks have become the state-of-the-art tool provide accurate, consistent fast data processing. However, published algorithms mostly solve only one specific problem they often require expert skills a considerable computer science machine background for application. ResultsWe thus developed deep pipeline called InstantDL four common processing tasks: semantic segmentation, instance pixel-wise regression classification. enables experts non-experts apply biomedical minimal effort. To make robust, we automated standardized workflows extensively tested it in different scenarios. Moreover, allows assess uncertainty of predictions. We benchmarked on seven publicly available datasets achieving competitive performance without any parameter tuning. For customization tasks, all code is easily accessible. Availability ImplementationInstantDL under terms MIT licence. It can be found GitHub: https://github.com/marrlab/InstantDL Contactcarsten.marr@helmholtz-muenchen.de