I have been racking my brain on this for few days. So please help me with these following queries:
What is the need to transform ray origins and ray directions in NDC frame ? I have already read https://github.com/bmild/nerf/issues/18, but still did not follow through.
In the paper, I read that the purpose for this is to preserve parallel lines and to have linear disparity-z relationship. But, the latter , though helps in better sampling and faster calculations, creates an inaccurate 3d space reconstruction. So, why is this desired ?
Why is NDC rays used in NeRF but not in TinyNeRF ?
What does "forward facing scenes" in the paper mean ? What does this have to do with NDC ?
I have been racking my brain on this for few days. So please help me with these following queries:
What is the need to transform ray origins and ray directions in NDC frame ? I have already read https://github.com/bmild/nerf/issues/18, but still did not follow through.
In the paper, I read that the purpose for this is to preserve parallel lines and to have linear disparity-z relationship. But, the latter , though helps in better sampling and faster calculations, creates an inaccurate 3d space reconstruction. So, why is this desired ?
Why is NDC rays used in NeRF but not in TinyNeRF ?
What does "forward facing scenes" in the paper mean ? What does this have to do with NDC ?