[Question / Feature Request] Comparisons with more recent NeRFs?

[xcyan]

Great work! I am curious if the authors have ever compared the Deblur-NeRF with more recent NeRFs (e.g., MipNeRF, MipNeRF360) where motion blur is handled implicitly (by rendering conical frustums instead of individual rays)? Some clarifications would be preferred.

[limacv]

Thank you for your inspiring question. Rendering using conical frustums as in MipNeRF and MipNeRF tries to solve the problem of multi-scale rendering (or anti-aliasing) instead of modeling blurriness. So they inherently cannot handle motion blur. Actually, this multi-ray-based rendering used in DeblurNeRF can be easily integrated into the mipNeRF (i.e. rendering and combining multiple frustums). This could be an interesting potential future work : )