How to use `DenseMatching` and `SingleImageDensePoseEstimation`?

[HtutLynn]

Hi, I am trying to benchmark lamar pipeline on CAB dataset using different feature extractors and matchers. However, I haven't found a way to do benchmarking with dense matching models like ROMA or LOFTR. I found LOFTR based implementation on dense_matching.py and pose_estimation.py but no documentation on how to use it. May I have some guidance on how to use it?

[mihaidusmanu]

As mentioned in the other issue, we currently do not support dense SfM in main. You can try #85 (I tested with LoFTR, other dense matchers would need to be integrated in hloc first).

The implementation that is in main uses lifting using the pseudo-GT mesh, not triangulation. It cannot be used with the benchmark data.

Closing as duplicate of #76