Euclidean space to be converted to NDC space

[bbzh]

Hi,

According to Eq.(16)-(17) in the NDC writeup and implementation in here, the Euclidean space to be converted to NDC space is chosen as the view frustum (or field of view) of the "averaged" camera pose. However, this does not guarantee to cover the union of view frustum formed by all the frames. Is there a reason not to choose a larger view frustum and hence larger Euclidean space?

Thanks

[kwea123]

In practice for LLFF dataset, the space viewed by other frustums might slightly go over the averaged frustum, but only a little (e.g. ~+-1.2), so it doesn't pose any problem on the positional embedding later.

I think you can choose a larger view frustum if indeed your poses are very different one from the other.