SynSin: End-to-end View Synthesis from a Single Image: Difference between revisions
No edit summary |
No edit summary |
||
| Line 1: | Line 1: | ||
SynSin: End-to-end View Synthesis from a Single Image (CVPR 2020) | SynSin: End-to-end View Synthesis from a Single Image (CVPR 2020) | ||
Authors: Olivia Wiles, Georgia Gkioxari, Richard Szeliski, Justin Johnson | Authors: Olivia Wiles, Georgia Gkioxari, Richard Szeliski, Justin Johnson | ||
Affiliations: University of Oxford, Facebook AI Research, Facebook, University of Michigan | Affiliations: University of Oxford, Facebook AI Research, Facebook, University of Michigan | ||
Revision as of 13:55, 3 August 2020
SynSin: End-to-end View Synthesis from a Single Image (CVPR 2020)
Authors: Olivia Wiles, Georgia Gkioxari, Richard Szeliski, Justin Johnson
Affiliations: University of Oxford, Facebook AI Research, Facebook, University of Michigan
Method
- First a depth map and a set of features are generated for each pixel using depth network \(d\) and feature network \(f\).
- The depths are used to create a 3D point cloud of features \(P\).
- Features are repositioned using the transformation matrix T.
- Repositioned features are rendered using a neural point cloud renderer.
- Rendered features are passed through a refinement network \(g\).