Kunhao-Liu / StyleRF

[CVPR 2023] StyleRF: Zero-shot 3D Style Transfer of Neural Radiance Fields
https://kunhao-liu.github.io/StyleRF/
143 stars 11 forks source link

Question for the zero-shot? #26

Closed SYSUykLin closed 7 months ago

SYSUykLin commented 7 months ago

May I ask where the zero-shot of your paper is presented? StyleRF, like other papers, is a two-stage training. I'm very confused about the zero-shot. Thanks

Kunhao-Liu commented 7 months ago

Hi, StyleRF is zero-shot to styles, meaning it does not need to be retrained for unseen new styles. Other method like ARF, styliednerf, snerf, need to be retrained for new styles.

SYSUykLin commented 7 months ago

Thanks for your reply, but I saw your paper need to train. In section 3: image

Kunhao-Liu commented 7 months ago

After this training stage, StyleRF does not need to be trained again for new styles, while other methods need to be retrained.

StyleRF can transfer styles it does not encounter during training.