I have many EM images to go through and segment out the mitochondria and then eventually analyze for area, perimeter,etc… which I can do with one image after training, however, I am trying to train a classifier to be used with all images so I can script the process. I can not seem to get a training classifier to properly segment the mitochondria in the various images I have. They are taken from different areas of the organ and some have a much darker/lighter background, some have a much darker/lighter nucleus…
I have been training using RandomForest classifier with an EM image and then saving the classifier and saving the Data.
Then open a different EM image and loading the saved classifier, then loading saved Data, then training again with another EM image. However, is this erasing the previous training I have been doing every time I train with a new EM image?
In case this makes a difference, I have opened all the original .dm4 files and then Downsampled all from 3600x2672 to 1800x1336 then saved as a .tif file in a new folder using a macro (which I was able to create thanks to these forums-thanks!!!)
I would love to be able to have a weka trained classifier to segment out the mitochondria in all my images without having to train again for each image.
Thanks for any help!