[WIP] tracksdata backend #126
Conversation
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## v2-dev #126 +/- ##
=========================================
Coverage ? 92.28%
=========================================
Files ? 50
Lines ? 2721
Branches ? 0
=========================================
Hits ? 2511
Misses ? 210
Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
🎉 wohoo! 🎉 |
…eature, removed all nx, made tests easier to read
|
PASSING! ✅🥳 @cmalinmayor, feel free to review the code and ask any questions regarding the many many changes necessary to make the backend |
cmalinmayor
left a comment
cmalinmayor
left a comment
There was a problem hiding this comment.
Not even close to done but the annotators action update I mentioned probably applies to other actions so I submit these now.
cmalinmayor
left a comment
cmalinmayor
left a comment
There was a problem hiding this comment.
Still didn't go through the tests thoroughly but lots of code comments to be getting on with! :)
| """ | ||
| lineages = nx.weakly_connected_components(self.tracks.graph) | ||
| max_id, ids_to_nodes = self._assign_ids(lineages, self.lineage_key) | ||
| lineages_internal = rx.weakly_connected_components(self.tracks.graph.rx_graph) |
There was a problem hiding this comment.
This whole section probably should be refactored to work better with the trackdata API. The internal external IDs seems fragile - this is a place where if we used tracksdata with SQL vs with rustworkx backend it would be a different implementation, right? Also, the _assign_ids helper could be refactored to do the efficient bulk update and then used for both lineage and track ids. There might be a better way to get the tracklet IDs that would eliminate the need for graph copying too.
| self.actions.append( | ||
| UpdateNodeSeg(tracks, new_value, all_pixels, added=True) | ||
| ) | ||
| tracks.graph[new_value][tracks.features.tracklet_key] = current_track_id |
There was a problem hiding this comment.
Hmm this should be handled by the tracksAnnotator, right? I'm less confident this time but I'm suspicious. I can look closer later if needed
| assert np.array_equal(mask1.mask, mask2.mask) | ||
|
|
||
|
|
||
| def compute_node_attrs_from_masks( |
There was a problem hiding this comment.
Ahh I think this is what the update function in the Annotators should do.
replaced nx in conftest and worked on test regarding adding/deleting nodes/edges