BIAFLOWS call for contribution (annotated datasets and BIA workflows)

Dear bioimaging community,

We are glad to announce the recent release of BIAFLOWS sandbox server (https://biaflows-sandbox.neubias.org) and the publication of BIAFLOWS in Patterns (Cell Press): https://doi.org/10.1016/j.patter.2020.100040.

We believe that openly benchmarking bioimage analysis workflows on representative annotated datasets is the best way to validate, disseminate, and further improve state-of-the-art algorithms, and we would like to promote BIAFLOWS as a central hub for this purpose.

To further populate BIAFLOWS, we would like to take this opportunity to launch a community call both for annotated microscopy datasets and for BIA workflows. BIAFLOWS currently support these classes of problems: object/landmark detection, object segmentation, particle/object tracking, pixel classification, filament tree/loopy network tracing. We are primarily interested in increasing the content for these classes of problems but we will also consider including datasets representative of other classes of problems when relevant.

Contribute with annotated datasets

We suggest to first post a minimal dataset description (problem illustrated, link to the dataset, number of images, dimensionality of the images, total size) to this thread: https://forum.image.sc/t/biaflows-datasets-contribution. We will then come back to you to initiate the process of aggregating the dataset to BIAFLOWS servers. Datasets for which image analysis workflows have already been developed (for instance in the context of a scientific publication or challenge) are the best target, but we are open to any contributions.

If you wish to further help us you can check if the dataset and annotations are compliant to the simple formats we specified here: https://neubias-wg5.github.io/problem_class_ground_truth.html, and convert them if necessary. If you are the owner, you should communicate which license you want to apply to the dataset (we highly encourage to use Creative Commons license when possible), or contact the owner to check whether the dataset could be aggregated to BIAFLOWS.

Contribute with Bioimage analysis workflows

First, you should identify which BIA problem you want to address (browsing https://biaflows.neubias.org/#/projects) and download the associated dataset to optimize your workflow for these images. Next, follow the guide https://neubias-wg5.github.io/creating_bia_workflow_and_adding_to_biaflows_instance.html to package your workflow for BIAFLOWS and test it on the sandbox server (http://biaflows-sandbox.neubias.org).

Alternatively, if you plan to process your own annotated dataset, upload it first to the sandbox server https://neubias-wg5.github.io/installing_populating_biaflows_locally.html. Once your workflow is running on the sandbox server (or if you need assistance), please notify us through this specific thread on image.sc: https://forum.image.sc/t/biaflows-workflows-contribution.

Thanks a lot for your interest in BIAFLOWS, and we are hoping to see you becoming an active contributor! Please contact us through the forum if you have specific questions.

The BIAFLOWS development team.

3 Likes