We propose pix2pix3D, a 3D-aware conditional generative model for controllable photorealistic image synthesis. Given a 2D label map, such as a segmentation or edge map, our model learns to synthesize a corresponding image from different viewpoints. To enable explicit 3D user control, we extend ...
In this work, we propose NerfDiff, which addresses this issue by distilling the knowledge of a 3D-aware conditional diffusion model (CDM) into NeRF through synthesizing and refining a set of virtual views at test time. We further propose a novel NeRF-guided distillation algorithm that ...
Another line of research explored 3D-aware portrait relighting, which leveraged the recent advances in un- conditional 3D-aware portrait generation [5] by combin- ing GANs [17] and NeRFs [27]. Concurrently, Jiang et al. [20] and Ranjan et al. [35] modeled the lighting ef- fects in ...
Conditional image-to-video gener- ation with latent flow diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18444–18455, 2023. 1 [30] Albert Pumarola, Antonio Agudo, Alberto Sanfeliu, and Francesc Moreno-Noguer. Unsupervised person image...
@click.option('--cond', help='Train conditional model', metavar='BOOL', type=bool, default=True, show_default=True) ... # Misc hyperparameters. @click.option('--p', help='Probability for --aug=fixed', metavar='FLOAT', type=click.FloatRange(min=0, max=1), default=0.2, show_...
At the point sampling stage, a point diffusion model learns the conditional distribution of point clouds given the input image. This stage is com- putationally efficient given the low resolution of the point clouds. The regression-based meshing stage transforms the sampled point...
Extensive evaluation show that our approach goes beyond competing conditional generators both in the capability to synthesize a much wider range of expressions ruled by anatomically feasible muscle movements, as in the capacity of dealing with images in the wild. 展开 ...