DL4MicEverywhere: deep learning for microscopy made flexible, shareable and reproducible

Abstract

DL4MicEverywhere is a platform that lets users train and implement their models in different computational environments. These environments include Google Colab, personal computational resources such as a desktop or laptop, and HPC systems. This flexibility is achieved by encapsulating each deep learning technique in an interactive Jupyter Notebook within a Docker container, enabling others to replicate analyses consistently across multiple platforms. DL4MicEverywhere (https://github.com/HenriquesLab/DL4MicEverywhere) enables users to install and interact with a large offering of standardized, user-friendly deep learning workflows, away from the limitations of proprietary platforms such as Google Colab and in a secure computational environment with controlled data privacy and resources. DL4MicEverywhere can be launched graphically, via X11 forwarding, or directly through a command line (headless mode), supporting HPC usage. This cross-platform containerization technology boosts the long-term platform’s sustainability and reproducibility, enhancing user convenience.

Ignacio Arganda-Carreras

Ikerbasque Research Associate Professor

My research interests include image processing, computer vision, and deep learning for biomedical applications.