Open Eisoc opened 6 days ago
Hi @Eisoc, sorry for the late response!
Yes, this function requires the input models to be trainer.models.values().
I've implemented some export_gaussians_to_ply
functions in both misc.py
and some Gaussian model files. These were used when I was debugging the code, and I haven't modified them to export standard .ply
yet.
I saw your PR, I'll look into it today, thanks for the contribution!
Hi ! I followed the guidelines very well and ran the code on some waymo datasets and was very shocked at the results. For better visualization, I think, exporting to ply is very good. And I see that you have already written this function.
https://github.com/ziyc/drivestudio/blob/388642a6a998833cb388e0f1a65dfc2071cc6a61/utils/misc.py#L50
I would like to know, does this function require the incoming models to be trainer.models.values(), and if so, maybe we have to put some getters inside the Gaussian model, such as vanilla, in order to read attributes such as self._means.