Awesome
DL4MicEverywhere
DL4MicEverywhere is a platform that offers researchers an easy-to-use gateway to cutting-edge deep learning techniques for bioimage analysis. It features interactive Jupyter notebooks with user-friendly graphical interfaces that require no coding skills. The platform utilizes Docker containers to ensure portability and reproducibility, guaranteeing smooth operation across various computing environments.
DL4MicEverywhere extends the capabilities of ZeroCostDL4Mic by allowing the execution of notebooks either locally on personal devices like laptops or remotely on diverse computing platforms, including workstations, high-performance computing (HPC), and cloud-based systems. It currently incorporates numerous pre-existing ZeroCostDL4Mic notebooks for tasks such as segmentation, reconstruction, and image translation.
Key Features
- 25+ Jupyter notebooks with a user-friendly graphical interface that requires no coding (scaling to 28+ soon).
- Docker-based packaging for enhanced portability and reproducibility.
- Deploys the ZeroCostDL4Mic user experience into local use.
- Supports a wide array of microscopy analysis tasks, including segmentation, reconstruction, registration, denoising, and more.
- Compatible with various computing environments, including laptops, workstations, HPC, and cloud with Docker.
- Automated build testing and versioning for improved reliability.
- Watch a short description.
Key benefits of DL4MicEverywhere
- Flexibility: Notebooks can run locally, in the cloud, or on high-performance computing infrastructure. No vendor lock-in.
- Reproducibility: Docker containers encapsulate the full software environment. Explicit versioning maintains stability.
- Transparency: Notebooks and models can be readily shared to enable replication of analyses.
- Accessibility: Interactive widgets and automated build pipelines lower barriers for non-experts.
- Interoperability: Adheres to data standards like BioImage Model Zoo for model sharing.
- Extensibility: Automated testing and Docker building streamline the addition of new methods.
DL4MicEverywhere is designed to make deep learning more accessible, transparent, and participatory. This enables broader adoption of advanced techniques while enhancing reliability and customization.
What is a DL4MicEverywhere notebook?
- A notebook to assist researchers in utilizing deep-learning models for image processing.
- It is fully encapsulated within a Docker container, providing a controlled and versioned snapshot of the dependencies required for the notebook.
- The versions of the required libraries are controlled upstream and downstream of the Docker container.
- The notebook is validated using continuous integration (CI) workflows to ensure compatibility with MacOS, Windows, and Linux.
- It features a user-friendly interface similar to ZeroCostDL4Mic, allowing users to train the model, perform quality checks, and run inference on new data.
- The notebook is fully traceable and open source.
Quickstart MacOs/Linux/Windows
- Download the ZIP file of the DL4MicEverywhere repository here and unzip it.
- Double-click the launcher in the DL4MicEverywhere folder that has the same name as your system (e.g.,
Windows_launch
for Windows operating systems). If this is the first time you run DL4MicEverywhere, we recommend you to follow the provided steps. - A GUI will automatically pop up. Choose a notebook and run!
With Docker, all dependencies are neatly bundled. Just launch and access deep learning workflows through an intuitive interface!
Refer to the Step-by-step "How to" guide and Requirements Installation Guidelines for further details.
Reproduce the demo in the video with the U-Net (2D) multilabel
notebook and Bacillus subtilis segmentation data from DeepBacs. Note that run time will vary from minutes to hours depending on the GPU availability and computing resources.
Requirements
DL4MicEverywhere rely on the following external software that is automatically installed when launching the tool.
If GPU acceleration is desired, the following needs to be installed:
- NVIDIA GPU + CUDA drivers (setup).
Contributing
We welcome contributions! Please check out the contributing guidelines to get started.
Documentation
- Step-by-step "How to" guide
- Remote Connection
- Notebooks
- Docker Desktop
- DL4MicEverywhere Technical Design
- Troubleshooting
- FAQ
- ZeroCostDL4Mic and DL4MicEverywhere over time
- Contributing Guidelines
Don't hesitate to reach out if you need any clarification!
Acknowledgements
We extend our gratitude to the ZeroCostDL4Mic contributors for their work on the original notebooks. We also thank the AI4Life consortium for their support and continuous feedback.
How to cite this work
Iván Hidalgo-Cenalmor, Joanna W. Pylvänäinen, Mariana G. Ferreira, Craig T. Russell, Alon Saguy, Ignacio Arganda-Carreras, Yoav Shechtman, AI4Life Horizon Europe Program Consortium, Guillaume Jacquemet, Ricardo Henriques & Estibaliz Gómez-de-Mariscal. DL4MicEverywhere: deep learning for microscopy made flexible, shareable and reproducible. Nat Methods 2024 DOI: https://doi.org/10.1038/s41592-024-02295-6
@article{hidalgo2024dl4miceverywhere,
title={DL4MicEverywhere: deep learning for microscopy made flexible, shareable and reproducible},
author={Hidalgo-Cenalmor, Iv{\'a}n and Pylv{\"a}n{\"a}inen, Joanna W and G. Ferreira, Mariana and Russell, Craig T and Saguy, Alon and Arganda-Carreras, Ignacio and Shechtman, Yoav and Jacquemet, Guillaume and Henriques, Ricardo and others},
journal={Nature Methods},
pages={1--3},
year={2024},
publisher={Nature Publishing Group US New York}
}