Overview

[almarklein]

Progress

• Primary work was just pushed to main.
• Tracker logic to e.g. connect with pygfx geometry in Connect to gfx mesh #1
• Some refactoring in Refactor higher level class #2 and Make example interactive via key events #3
• Started on todo in Add undo example #4
• More refactoring in Some basic refactoring #5, Refactor usage of checking internal state #6, Add test in pop_vertices #7, Use draw_range in pygfx buffer #8
• Setup an actual morph example in Morph example #9
• Better geodesic distance in Better distance over surface #10
• Refactor example to organize all objects better in Refactor example #12
• Smoothing in Implement brush-smoothing #13
• Resampling in Resampling #14
• Some renaming (names clash a bit in a few places).
• Morph using mesh slices in Update example to use mesh slice #18
• More realistic example data.
• Smoothing pointy objects? (can still be challenging, but should be considerably easier now)
• Improve performance of selecting the vertices? Faster selection #17

Possible future work

Things to definitely solve someday:

• Proposal for what to do with this code #20
• The resampling step can sometimes make the mesh non-manifold #23
• Improve buffer/texture update_range pygfx#600

Usability improvements:

• Show the slices (axii) in the 3D view.
• Show CT data together with mesh slices.
• Clicking next to the mesh slice still grabs it.
• Improve geodesic distance measuring by taking curvature into account.

Performance improvements:

• Cythonizing the mesh selection and resampling.
• Improve performance by enabling chunking in gfx Buffer.update_range.

Worth noting but likely not a bottleneck so may bot be worth the effort:

• Keeping track of normals could perhaps be delayed if we don't need it for the viz (i.e. no updating normals during the interaction).
• The _add_vertex_on_edge can likely be vectorized to some extent. This is much harder for _pop_vertex though, so Cython is likely a more reliable path towards better performance.