I was wondering how exactly the uncertainty maps (for example in the pixel classification), generated by Ilastik, are calculated? For a given voxel: does Ilastik simply divide the number of trees, voting for each other phase, by the number of trees voting for the phase that the voxel was segmented as? Here’s one example, if my explanation was unclear 10 Trees voted a voxel to belong to phase 1, 10 Trees voted for phase 2 and 80 Trees voted for phase 3. So the uncertainty would be 20/80 = 0.25.
Thanks in advance,