search

NVlabs / neuralangelo

Official implementation of "Neuralangelo: High-Fidelity Neural Surface Reconstruction" (CVPR 2023)
https://research.nvidia.com/labs/dir/neuralangelo/
Other
4.35k stars 388 forks source link

How to "run neuralangelo" #196

Open chlrhalsgur opened 6 months ago

chlrhalsgur commented 6 months ago

Hello. I'm doing this step

Run Neuralangelo! EXPERIMENT=toy_example GROUP=example_group NAME=example_name CONFIG=projects/neuralangelo/configs/custom/${EXPERIMENT}.yaml GPUS=1 # use >1 for multi-GPU training! torchrun --nproc_per_node=${GPUS} train.py \ --logdir=logs/${GROUP}/${NAME} \ --config=${CONFIG} \ --show_pbar

And im doing with custom data. I followed the steps below

  1. conda activate neuralangelo
  2. Data Preparation
    • git submodule update --init --recursive
    • bash projects/neuralangelo/scripts/preprocess.sh ${SEQUENCE} ${PATH_TO_VIDEO} ${DOWNSAMPLE_RATE} ${SCENE_TYPE}
    • bash projects/neuralangelo/scripts/preprocess_dtu.sh ${PATH_TO_DTU}
  3. torchrun --nproc_per_node=${GPUS} projects/neuralangelo/scripts/extract_mesh.py \ --config=${CONFIG} \ --checkpoint=${CHECKPOINT} \ --output_file=${OUTPUT_MESH} \ --resolution=${RESOLUTION} \ --block_res=${BLOCK_RES}

And at the last step, i got only this message "failed to create process."

Am i missing something?

And i have two more questions.

If i want to process just for my custom data, do i need other datasets? ex) tanks_and_temples, DTU And do i need to train the 'model'? Isn't it fine to do forward process?

Thank you!