Bioimage model zoo - a community-driven resource for accessible deep learning in bioimage analysis

Authors: Wei Ouyang, Fynn Beuttenmueller, Estibaliz Gómez-de-Mariscal, Constantin Pape, Tom Burke, Carlos Garcia-López-de-Haro, Craig Russell, Lucía Moya-Sans, Cristina de-la-Torre-Gutiérrez, Deborah Schmidt, Dominik Kutra, Maksim Novikov, Martin Weigert, Uwe Schmidt, Peter Bankhead, Guillaume Jacquemet, Daniel Sage, Ricardo Henriques, Arrate Muñoz-Barrutia, Emma Lundberg, Florian Jug, Anna Kreshuk
Technologies: BioImage Model Zoo and ZeroCostDL4Mic
Preprint published in bioRxiv, January 2022
Publisher: Cold Spring Harbor Laboratory

Bioimage model zoo - a community-driven resource for accessible deep learning in bioimage analysis
DOI: 10.1101/2022.06.07.495102

The manuscript introduces the BioImage Model Zoo, a community-driven resource for sharing, exploring, testing, and downloading standardized pre-trained models for deep learning (DL) in bioimage analysis. The Zoo aims to make DL methods for microscopy imaging findable, accessible, interoperable, and reusable (FAIR) across software tools and platforms. It supports a unified way of describing and consuming trained DL models through a standard model description format and libraries for standardized model execution. The Zoo is already supported by several tools, including ilastik, deepImageJ, ImJoy, StarDist, ZeroCostDL4Mic, and QuPath, and is open to new community partners. The manuscript also discusses the development and implementation of various tools and libraries for handling and analyzing bioimage data using DL techniques, such as bioimageio.core, core-bioimage-io-java, and imagej-modelzoo. The Zoo provides a user-friendly web application, BioEngine, for quick evaluation of models and supports interoperability and reproducibility. It is expected to promote rapid dissemination of DL developments, enrich and support advanced bioimage analysis workflows, and adhere to FAIR principles.