Differentiable version of the box3d_overlap function

facebookresearch / pytorch3d

PyTorch3D is FAIR's library of reusable components for deep learning with 3D data

https://pytorch3d.org/

Other

8.66k stars 1.3k forks source link

Differentiable version of the box3d_overlap function #994

Open lioutasb opened 2 years ago

lioutasb commented 2 years ago

🚀 Feature

It would be extremely useful to add gradient support for the box3d_overlap function which will allow being used as an objective function. I was wondering if there are plans to support this in the near future.

nikhilaravi commented 2 years ago

@lioutasb great to hear you are using this feature! We have been thinking about this but have not yet finalized our roadmap for 2022.

themantalope commented 2 years ago

This would be really beneficial for my projects as well.

GregorKobsik commented 3 months ago

I would also like to have the box3d_overlap function implemented with support for gradient computation. Furthermore, I am willing to put effort to implement it on my own.

@nikhilaravi @gkioxari Is there any chance to get some help or supervision in implementing this feature?