NVlabs / neuralangelo

Official implementation of "Neuralangelo: High-Fidelity Neural Surface Reconstruction" (CVPR 2023)
https://research.nvidia.com/labs/dir/neuralangelo/
Other
4.27k stars 383 forks source link

Can I reduce Polygon count? #181

Closed MicKkmack closed 4 months ago

MicKkmack commented 5 months ago

Hi ,

I've managed to get some good results , but I'm struggling to load them due to the amount of polygons.

I'm getting around 15,000,000 polygons, whereas Nerf Studio gives me around 50,000 , does anyone know if there is a parameter to lower this or what causes that many polygons?

Thanks! :)

Bombaninha commented 4 months ago

Lower RESOLUTION to reduce the mesh size during the isosurface extraction stage:

CHECKPOINT=logs/${GROUP}/${NAME}/xxx.pt
OUTPUT_MESH=xxx.ply
CONFIG=logs/${GROUP}/${NAME}/config.yaml
RESOLUTION=2048
BLOCK_RES=128
GPUS=1  # use >1 for multi-GPU mesh extraction
torchrun --nproc_per_node=${GPUS} projects/neuralangelo/scripts/extract_mesh.py \
    --config=${CONFIG} \
    --checkpoint=${CHECKPOINT} \
    --output_file=${OUTPUT_MESH} \
    --resolution=${RESOLUTION} \
    --block_res=${BLOCK_RES}
MicKkmack commented 4 months ago

Lower RESOLUTION to reduce the mesh size during the isosurface extraction stage:

CHECKPOINT=logs/${GROUP}/${NAME}/xxx.pt
OUTPUT_MESH=xxx.ply
CONFIG=logs/${GROUP}/${NAME}/config.yaml
RESOLUTION=2048
BLOCK_RES=128
GPUS=1  # use >1 for multi-GPU mesh extraction
torchrun --nproc_per_node=${GPUS} projects/neuralangelo/scripts/extract_mesh.py \
    --config=${CONFIG} \
    --checkpoint=${CHECKPOINT} \
    --output_file=${OUTPUT_MESH} \
    --resolution=${RESOLUTION} \
    --block_res=${BLOCK_RES}

Thanks a lot, that worked :)

camolina2 commented 1 month ago

Hi,

I was wondering if there is a way to extract the mesh in parts. I am running out of GPU memory for the extraction, and I would like to achieve a better resolution.

Thank you!