Nice work! When reading your paper, I noticed that your training data is generated by "rotating around the vertical-axis" with plastimatch, which is more like CBCT reconstruction. However, x-ray should be possibly from any 3D views.
While the training data from the original NeRF also contains different views besides rotating around the vertical-axis, I wonder whether it is possible to generate data with more diverse view (arbitrary 3D views) with plastimatch.
It will be appreciated if you have any suggestions.
Nice work! When reading your paper, I noticed that your training data is generated by "rotating around the vertical-axis" with plastimatch, which is more like CBCT reconstruction. However, x-ray should be possibly from any 3D views.
While the training data from the original NeRF also contains different views besides rotating around the vertical-axis, I wonder whether it is possible to generate data with more diverse view (arbitrary 3D views) with plastimatch.
It will be appreciated if you have any suggestions.