Include phase cross correlation flow

[lmanan]

Proposed Change

This PR primarily includes calculation of flow by phase cross correlation and corrects the edge distance feature by incorporating the now available flow. It is designed to be used as follows:

from motile_toolbox.candidate_graph.compute_graph import get_candidate_graph
from motile_toolbox.utils.flow import compute_pcc_flow, correct_edge_distance

# Load segmentation and raw image data
segmentation = ...
raw = ...

# Create candidate graph, initially ignoring flow
candidate_graph, conflict_ids = get_candidate_graph(segmentation=segmentation,
                                                    max_edge_distance=50,
                                                    iou= True)

# Calculate phase cross correlation flow 
compute_pcc_flow(candidate_graph, images=raw)

# Correct previously calculated edge distance 
candidate_graph = correct_edge_distance(candidate_graph)

#  Execute motile-related commands, as usual
track_graph = TrackGraph(candidate_graph)
...

By calculating the flow separately and after constructing the candidate graph, allows the flexibility to replace this step by a flow coming from an alternate source (for example, a DL model).

Checklist

Go through these things before merge. Actions should run automatically to test them, but for information on how to run locally, see CONTRIBUTING.md.

• [] I have added tests that prove that my feature works in various situations or tests the bugfix (if applicable).
• [] I have checked that the tests pass and I maintained or improved test coverage (if applicable).
• I have written docstrings and checked that they render correctly in the documentation build.
• I have checked that mypy type checking passes.

Further Comments

Other minor changes include:

1. Specifying the python version to be 3.10 onwards (f75e60d) since the typing does not work for versions below.
2. Obtaining motile from the github (6c27a49).
3. Adding BBOX and FLOW by default (210371f).

[cmalinmayor]

Main change needed is to make a new edge attribute to hold the new value, something like FLOW_DISTANCE, instead of overwriting the distance attribute. A couple other nitpicks.

[cmalinmayor]

unpack into node_id, data would be nicer to read

[cmalinmayor]

Can get rid of loc and use the size of the bbox to infer the number of dimensions

[cmalinmayor]

What does this represent? Is it a single number? A vector? A vector at each pixel in the input image?

[cmalinmayor]

First, I strongly don't think this should overwrite distance, it needs its own attribute. Second, Jan and I removed the distance attribute (and updated Motile.EdgeDistance cost) 😆 so now this feature definitely needs its own attribute!

[cmalinmayor]

again, I prefer unpacking the tuple here rather than using [2] later

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 12.90323% with 27 lines in your changes are missing coverage. Please review.

Project coverage is 77.13%. Comparing base (4221b06) to head (570c976).
Report is 1 commits behind head on main.

Files	Patch %	Lines
src/motile_toolbox/utils/flow.py	0.00%	27 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #10      +/-   ##
==========================================
- Coverage   83.83%   77.13%   -6.71%     
==========================================
  Files          12       13       +1     
  Lines         297      328      +31     
==========================================
+ Hits          249      253       +4     
- Misses         48       75      +27     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.