add two SfM papers with symmetry and Manhattan constraints

[ethanygao]

Add two recent SfM papers which exploited symmetry and Manhattan constraints.

• They both proposed novel SfM methods for category-specific object 3D reconstruction.
• They both are global SfM.
• The ECCV 2016 paper proposed two novel non-rigid SfM algorithms.
• The CVPR 2017 paper proposed one novel single image algorithm and one rigid SfM algorithm.

[pmoulon]

Thank you for your PR. I don't think we can put your paper in the generic category, since your papers are adapted only for some category of objects. All the other paper are object agnostic.

Do we can find a category to put those paper in?

[ethanygao]

Hi Pierre, thanks for the discussion!

You are right that they both are applied for object reconstruction, where different object instances (with different viewpoints) from the same category are used as input with rigid or non-rigid deformation assumption between them. This indeed differs from the "generic" SfM which used a video of some same objects from different viewpoints as input.

But both papers also contribute to the "generic" SfM methods as novel SfM algorithms are proposed to exploit symmetry and Manhattan constraints (for both rigid and non-rigid cases), this can be also applied using videos as input.

Any suggestions to change their category?

[pmoulon]

What about adding a category like SfM with symmetry and Manhattan constraints or Constrained SfM?

It would be nice if you can add some other major references in this field too in order to guide the readers to the other trading work in this field.

[ethanygao]

Thanks! I think Constrained SfM is great, and have updated the PR accordingly. To the best of my knowledge, there is not much work in this field (SfM with additional priors), I will try to add them once I find more.