Closed jjxjiaxue closed 4 months ago
Hi! Thank you for your interest! To perform camera calibration, we do not use either depthmap or surface normal map. The depthmap and surface normal are proof of concept methods where the camera intrinsic is acquirable from the common 3D priors. However, this baseline requires an impractical precision for depth & surface normal. The calibration method benchmarked instead uses the incidence field (as an alternative to the depth & surface normal).
Thank you so much
Hi, thank you for your promising work! I have a few questions about the surface normals priors. Do they play a significant role in the model's performance? If I use a different monocular metric depth estimation model instead of the new-crfs model, can I still input that depth into the incidence head? Will it work just as effectively? Thank you so much.