SynSin End-to-end View Synthesis from a Single Image

a hs pc of learned features

• trained using image pairs and camera poses
• image -> feature and depth map -> lift to 3d -> project to novel view -> image
• biggan… not doable for me
• soft z-buffer and a disk of influence for each point
  • alpha compositing