If you are able to share an image collection publicly you may be able to
entice someone already working on this, to collaborate and help you.
Sometimes people are already working on things, but the test images are
private, sometimes they only have a small number of tests sets, and more
tests are helpful.
I am very interested in this subject too. It would be great to have deep
learning plugins in ImageJ/KNIME. There is already work underway. At some
point I'd like to get involved with this myself, but for now I am busy with
I believe most of the work will be in setting up a nice workflow and/or GUI
to annotate/organize/label the test images. It is fairly easy to set up a
Deep learning work flow, once you have the data organized.
One thing I would like to see, is a plugin collection that clearly
differentiates "object identification" and "pixel based clasiffier". If
you have a bunch of pictures of Cats and Dogs, and want to know which ones
have cats, and which ones dogs, there are a lot of deep learning tutorials
that show you how to set this up in a day.
The problem I deal with, much more often, is pixel based classification. I
have images of cells, and need to label individual pixels that belong to
different components of the cell.
On Fri, Jul 21, 2017 at 4:23 PM, Ankur Kaushal <[hidden email]>