Server Analysis API

[spli]

Aim

Create an API so that external analysis algorithms can be integrated with OMERO.

Current

There is a scripting API but it is designed for relatively small jobs. This requirement is for an API to handle large computationally intensive jobs over many images and datasets.

Scope

This is intended to cover the backend components such as running jobs and handling storage of intermediate and final results.

There will be some user-interface development to visualise results, but any major visualisation work should probably form a separate requirement.

OMERO.searcher and WND-CHRM will be used to develop and evaluate the API, and will also form reference implementations/examples.

Workflow

The general workflow from a user's point of view is as follows:

1. Preprocess images #9947
2. Calculate image features, store them #9948
3. Train an algorithm to understand images #9949
4. Use the trained algorithm to make predictions or search images #9951
5. View the results #9952

Depending on the algorithm some of these steps will be transparent to a user. For example, a general search application such as OMERO.searcher may continuously run in the background, automatically calculating features and updating its knowledge, so a user would only be exposed to the search interface.

A data-exploration tool might effectively build any machine learning into the visualisation stage for real-time exploration. This means only the feature extraction and visualisation steps are required.

The context of a learning algorithm (#9950) will require some thought to make it usable by non-specialists.

See also

• #7902 Message queue support, this would be useful for resource-intensive calculations on large datasets where a cluster is available.
• #1856 Scripting improvements.