search

NVlabs / neuralangelo

Official implementation of "Neuralangelo: High-Fidelity Neural Surface Reconstruction" (CVPR 2023)
https://research.nvidia.com/labs/dir/neuralangelo/
Other
4.4k stars 390 forks source link

switch to analytic gradients after delta is small enough? #125

Closed blacksino closed 1 year ago

blacksino commented 1 year ago

would this bring any performance degradation？ or may speed up training a bit?

mli0603 commented 1 year ago

Hi @blacksino

Yes, it will speed up the training slightly! Please find the detailed speed comparison in Table 5 of the Appendix.