[ome-devel] OMERO analysis, image search and alternative storage
s.p.li at dundee.ac.uk
Mon Nov 26 15:46:54 GMT 2012
I've recently started as a new OME developer in Dundee, and I'll be working on integrating analysis and machine learning applications, such as image search, into OMERO. Many of you will have already come across OMERO.searcher from Robert Murphy's group and WND-CHRM from Ilya Goldberg's group, and hopefully they'll be integrated into OMERO over the next few months.
In general there are two main aims:
1. Make image search and automated classification accessible to all OMERO users.
2. Define an OMERO analysis and machine learning API so that others can easily modify their own algorithms to work with OMERO without having to duplicate effort.
I think the latter can be split into 3 main areas:
* Feature extraction (How will algorithms/scripts be run, how will the features be stored?)
* Training (Again how will scripts be run, how will parameters be handled, etc)
* Testing/Usage (How will someone choose the right algorithm?)
The first thing I'm looking at is alternative storage for features. The current OMERO.tables is a wrapper around HDF5, but doesn't take advantage of it's full capabilities, and requires the numbers of features to be defined in advance which isn't ideal for everyone. A NoSQL database such as MongoDB is another potential option that's been under consideration for a while.
If anyone has any comments, either for machine learning and image search in general, or on the subject of alternative storage, I'd love to hear from you, especially if you have any example use-cases you'd like to share.
The University of Dundee is a registered Scottish Charity, No: SC015096
More information about the ome-devel