search

google / ffn

Flood-Filling Networks for instance segmentation in 3d volumes.
Apache License 2.0
317 stars 100 forks source link

What about consensus and agglomeration steps? #20

Open subeeshvasu opened 4 years ago

subeeshvasu commented 4 years ago

Hello,

Can someone please explain about the codes corresponding to the steps consensus and agglomeration (as in the nature paper) that follows the inference. https://github.com/google/ffn/blob/master/ffn/inference/consensus.py and https://github.com/google/ffn/blob/master/ffn/inference/resegmentation.py seems to have routines associated with these two steps. However, there is no instruction about the usage of these modules. Has anyone succeeded in running these two steps to yield the so called output ffn-c?

Thanks,

oradomskyi commented 4 years ago

Hi @subeeshvasu, if the question is still actual for you: recently I managed to test both consensus and resegmentation.

in the heart of consensus is boolean AND comparison between two segmentations - A and B. First, if I understood properly, largest overlaping segment pairs are selected, then A being updated with new segments IDs at all pixels where two masks intersect. there is a possibility to filter out result items smaller than N voxels.

resegmentation, as far as I can tell, uses prior segmentation and prior probabilities and running segmentation of a single item starting at explicitly defined location and restricted in a certain area.

Hope this helps, Alex