Today I discovered the marvel that is Bio7. This eclipse-like IDE combines R and ImageJ and their website has a tutorial on how to do image feature extraction and ML with this combination.
I've followed through the tutorial, but instead of trying to classify parts of a single image, I'm dealing with > 100,000 images.
I've found that I can get conformable data for test and control but only if I do not select ROI's in the test data. While the data objects -- once passed to R and transformed with an appropriately specified as.data.frame(matrix(unlist()))
statement, I have to wonder if I'm not losing a lot of value in terms of classification by not indicating the ROI's.
How can I get an R object that has 1 element (i.e. 1 matrix in a list of matrices, or whatever) per image (or per frame/slice of an Image Sequence) after specifying the ROI in the ROI manager?
See also on YouTube:
https://www.youtube.com/watch?v=CyGB8uUjbWk&list=PLzCgXMp4TBsVKSRHZ8Q9y_3ZSlGl2kcrp&index=3
Apropos the 'Selection' button only transfers the selection coordinates and not the pixel data.
See: