normalize UV coordinates to [0,1] in make_material_atlas()

[Colin97]

When building a material atlas, it's common to have negative UV coordinates (e.g., Shapenet dataset).
Existing options (texture_wrap="repeat", texture_wrap="clamp", texture_map=None) may fail to produce correct textures.
Failure case 1: texture_wrap = "repeat"

Failure case 2: texture_wrap = "clamp"

Failure case 3: texture_wrap = None

However, with this commit, using "texture_wrap = None" can produce correct textures for both cases:

[facebook-github-bot]

Hi @Colin97!

Thank you for your pull request and welcome to our community.

Action Required

In order to merge any pull request (code, docs, etc.), we require contributors to sign our Contributor License Agreement, and we don't seem to have one on file for you.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!

[facebook-github-bot]

Thank you for signing our Contributor License Agreement. We can now accept your code for this (and any) Meta Open Source project. Thanks!

[bottler]

As you indicate, UV-coordinate wrapping is a feature which PyTorch3D should support, and this PR indicates part of the code which needs to be changed. Unfortunately I think a proper solution is quite complicated. IIUC, UV coordinates should be understood as in an infinite plane which wraps around (topologically a torus). So in general, a single face can encompass parts from many copies of the original map. TexturesUV.join_scene would become much more complicated.

Another enhancement to the class which seems to be wanted is having multiple texture maps instead of one (per batch element). Implementing that would make this easier.

[Colin97]

@bottler Yes, TexturesUV.join_scene can be very complicated.

However, the proposed PR is about TexturesAtlas not TexturesUV. Specifically, when creating a texture atlas from a single image texture, the above change ensure a correct sampling and thus creating a correct atlas. In this case, we already calculated all RxR UV coordinates for each atlas pixel (which may be negative) and need to sample from the image. The proposed wrapping is thus proper and necessary .

[bottler]

Yes, sorry, I was confused about which part this was changing. Let's continue the discussion for the atlas on #1270.