search

janelia-flyem / gala

Automatic segmentation of electron microscopy volumes
BSD 3-Clause "New" or "Revised" License
76 stars 29 forks source link

Tree-based agglomeration #32

Closed jni closed 9 years ago

jni commented 10 years ago

I've been thinking about optimizing gala. I think one of the first steps is to stop merging segments during agglomeration, and just keep track of the merges in a tree. I've created a Python tree library viridis (after Morelia viridis, the green tree python) and use it to keep track of the merge tree. All the segment merging logic is gone, simplifying the codebase while increasing efficiency. It passes all the current tests (admittedly few, but they cover the essentials of agglomeration and feature computation).

Please let me know if there are any objections to this change.

jni commented 10 years ago

Don't merge this! Turns out Travis was "passing" despite failing tests! (See #33.) Will rebase and fix failing tests.

coveralls commented 9 years ago

Coverage Status

Changes Unknown when pulling 740bebbc9fa793152c63980e135cfdde8eafeb3a on jni:tree-agglo into \ on janelia-flyem:master**.

jni commented 9 years ago

I'm merging this as a precursor to a release. Boom!