Some things we'd like to see discussed at the hackathon:
- Modularity
- Can the alignment pipeline be separated into independent steps with a well-defined format for the intermediate results? This would help in parallelization on the cluster, as well as easy experimentation with new approaches. It also would help separate control parameters into groups based on which step they control.
- 1D and 2D constraints
- Currently, filtering correspondences removes all of those except corner-like matches. Is it possible to classify correspondences into smooth/linear/corner? 1D correspondences can be used to constrain the solution to a subspace, and might result in improved results and solution times in some cases.
- Cracks
- Some of our data has (long, linear) cracks from the microtoming process. We can detect them automatically. It would be useful to be able to eliminate cracked regions from correspondence finding, and to be able to do some mesh surgery on the structural mesh during the solving pr