search

autonomousvision / sdfstudio

A Unified Framework for Surface Reconstruction
Apache License 2.0
1.9k stars 179 forks source link

is monosdf and neuralangelo inherently slow? #284

Closed hanjoonwon closed 5 months ago

hanjoonwon commented 5 months ago

image When I use neus,neuralangelo,monosdf,etc except neus-facto,the train iter is quite slow as shown below. Is this a normal result? My graphic card is rtx 2060 vram 6gb

i trained with ns-train neus --pipeline.datamanager.train-num-rays-per-batch 1024 --pipeline.model.sdf-field.bias 0.3 --pipeline.model.sdf-field.use-grid-feature False --pipeline.model.background-model mlp --pipeline.model.mono-depth-loss-mult 0.1 --pipeline.model.mono-normal-loss-mult 0.05 --vis wandb --pipeline.model.sdf-field.inside-outside False --trainer.steps_per_save 5000
--trainer.steps-per-eval-image 5000 --trainer.max-num-iterations 300000 --experiment-name newwvidbat sdfstudio-data --data /workspace/processdata/vidbat

niujinshuchong commented 5 months ago

Hi, it depends on the point sampling and the network architecture. NeuS is slower than neus-facto since more points are sampled on the ray and the default nextowrk architecture is different.

hanjoonwon commented 5 months ago

Hi, it depends on the point sampling and the network architecture. NeuS is slower than neus-facto since more points are sampled on the ray and the default nextowrk architecture is different.

@niujinshuchong in case of neuralangelo iter time 1 miniute.. that is because of my gpu?? I use rtx 2060