This is the implementation of NeurIPS 2022 paper Neural Transmitted Radiance Fields.
- In Eq.(7) and (8),
$k$ denotes the total number of views, instead of pixels (typo in an older version).
Input example (w/ reflection) | Ours, 6 training views. | Ground truth (w/o reflection) |
NeRF, 6 training views. | NeRF, 18 training views. | NeRF-W, 6 training views. |
[TODO] Coming soon!
Our data is available at https://drive.google.com/drive/folders/1Gwzz8DgORrd9b_VEdk9z2108qJXz6bfk