Neural scene representation and rendering (Science 2018): Difference between revisions
(Created page with "Neural scene representation and rendering (Science 2018) Authors: S. M. Ali Eslami*,†, View ORCID ProfileDanilo Jimenez Rezende†, View ORCID ProfileFrederic Besse, View O...") |
No edit summary |
||
| Line 11: | Line 11: | ||
* A representation network creates a neural representation \(\mathbf{r}\). | * A representation network creates a neural representation \(\mathbf{r}\). | ||
* The generation network predicts the scene from viewpoints \(\mathbf{v}^q\) and random noise \(\mathbf{z}\). | * The generation network predicts the scene from viewpoints \(\mathbf{v}^q\) and random noise \(\mathbf{z}\). | ||
[[Category: Papers]] | |||
Latest revision as of 16:39, 31 January 2022
Neural scene representation and rendering (Science 2018)
Authors: S. M. Ali Eslami*,†, View ORCID ProfileDanilo Jimenez Rezende†, View ORCID ProfileFrederic Besse, View ORCID ProfileFabio Viola, View ORCID ProfileAri S. Morcos, Marta Garnelo, View ORCID ProfileAvraham Ruderman, View ORCID ProfileAndrei A. Rusu, Ivo Danihelka, Karol Gregor, David P. Reichert, Lars Buesing, View ORCID ProfileTheophane Weber, Oriol Vinyals, Dan Rosenbaum, Neil Rabinowitz, Helen King, Chloe Hillier, View ORCID ProfileMatt Botvinick, Daan Wierstra, Koray Kavukcuoglu, Demis Hassabis
Affiliations: Deepmind
Generative Query Network
- We have K images for scene i: \(\mathbf{x}^k_i\) taken from 2D viewpoints \(\mathbf{v}^k_i\)
- A representation network creates a neural representation \(\mathbf{r}\).
- The generation network predicts the scene from viewpoints \(\mathbf{v}^q\) and random noise \(\mathbf{z}\).