Closed NagabhushanSN95 closed 1 year ago
Yes, they don't use coarse to fine network,because they use another strategy to optimize the sampling process , they use an occupied_grid to skip the empty region, which is mentioned in instant-ngp.
Okay, I'll check that. Thanks!
Hi,
I noticed that in train_mlp_nerf.py file, a NeRF MLP is queried only once with stratified sampling. Do you not use hierarchical sampling with coarse and fine MLPs?