iNeRF #371
Replies: 2 comments 2 replies
-
I'm interested in that. I was watching the code and I believe that it's possible, but don't know how yet. |
Beta Was this translation helpful? Give feedback.
-
Hi there, instant-ngp supports training of camera extrinsics (as well as an intrinsic distortion grid for that matter). You can simply tick the "Train extrinsics" checkbox in the GUI to enable it. If you want to first pre-train a model with known extrinsics and then reconstruct the pose from additional viewpoints, you can
Since this sequence of steps is rather cumbersome, I'd recommend using the Python bindings for this sort of stuff. These will also let you skip the checkpoint dance as they allow you to replace the dataset on-the-fly. Cheers! |
Beta Was this translation helpful? Give feedback.
-
Hi,
Thanks for sharing this research and repository! NeRF has become much more accessible opening up new research and use-cases.
Are there any plans to include iNeRF functionality to derive camera poses relative to a trained model? Having this key functionality easily accessible as well would open even more use-cases.
Either way thanks,
Roy
Beta Was this translation helpful? Give feedback.
All reactions