kiui

@skunkwerk Hi, yes, the pose file has an 'eye_ratio' entry for each frame [here](https://github.com/ashawkey/RAD-NeRF/blob/main/nerf/provider.py#L161), but currently we don't use it in the inference code.

@skunkwerk Which dataset are you using? Since the blinking is learned from data too, some identities may fail to learn how to depend it on `eye_ratio`.

This is strange, the obama dataset should be able to control the eyes well as in the demo video. Maybe you could try with the GUI first?

It seems the GPU arch is too old to build the CUDA extension.

@PLEXATIC Hi, thanks for your interest! 1. Yes, just put the `obj, mtl, png` files in the same folder. 2. The text-to-3D pipeline does encounter many unsatisfactory results. We are...

@kendo6666 Hi, it's not meant to be used. Basically it means we provide audio features as class labels and learn an embedding per class instead of directly passing in per...

@x-CK-x Hi, since our model is essentially a U-Net with attention like stable diffusion, it's possible to add LoRA. But it needs experiments to decide which layers to apply, and...

Thanks @Ayushlion8! @guwa1 Sorry that this repo only aims to perform single-view generation, you may check other methods for sparse view reconstruction, like this new paper today: https://hwjiang1510.github.io/LEAP/

Cite from the blog: > After training several MLPs with different numbers of layers and parameters and different activation functions, we found that a simple linear model on the top...

@mystijk Sorry this is because I changed the network but forget to update pretrained checkpoints. I will revert it soon.