PointNet++ lifting (pointcloud to hypergraph) + new dataset

[sukjulian]

Lifting

This lifting generates a hypergraph by mimicking the set abstraction layer of PointNet++. This is done by farthest point sampling and subsequent clustering of points within a radius of the sub-sampled points. These clusters then constitute hyperedges which can be used to pool information from local neighbourhoods. Constructing a complete PointNet++ from this would require us to recursively apply the lifting while regarding the previous hyperedges as the new nodes (probably currently not supported by the code base).

Dataset

PointNet++ is a popular model in the biomedical engineering community. Thus we implement a loader for a new dataset (described in this paper) consisting of 2000 synthetic coronary artery meshes with vertex-wise directional wall shear stress labels. The dataset can be regarded as simplicial complex in the sense that face features (surface normal) and an assignment of each vertex to its incident triangles are available. In the biomedical engineering community, datasets like these are usually treated as point clouds, by dropping the face connectivity. This is also what we do here to demonstrate our lifting. However, it would be interesting to open these datasets up to topological deep learning on simplicial complices.

Authors

Submission by team MIA-UT: Patryk Rygiel (@PatRyg99) | Julian Suk (@sukjulian)

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[gbg141]

Hello @sukjulian! Thank you for your submission. As we near the end of the challenge, I am collecting participant info for the purpose of selecting and announcing winners. Please email me (or have one member of your team email me) at guillermo_bernardez@ucsb.edu so I can share access to the voting form. In your email, please include:

• your first and last name (as well as any other team members)
• the title of the method you implemented
• the input domain of the method you implemented
• the output domain of the method you implemented
• your pull request number (#53)

Before July 12, make sure that your submission respects all Submission Requirements laid out on the challenge page. Any submission that fails to meet this criteria will be automatically disqualified.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 50.79365% with 31 lines in your changes missing coverage. Please review.

Please upload report for BASE (main@497947b). Learn more about missing BASE report.

Files	Patch %	Lines
modules/data/utils/utils.py	11.42%	31 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #53 +/- ## ======================================= Coverage ? 65.00% ======================================= Files ? 18 Lines ? 523 Branches ? 0 ======================================= Hits ? 340 Misses ? 183 Partials ? 0 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.