nerf-optimization

Containerized research environment for NeRF optimization.

Setup

Run the following before using docker compose to fix ownership while inside a development container:

$ chmod +x initialize.sh
$ ./initialize.sh

This only needs to be done once. Confirm that you're project is initialized by checking the .env file. It should contain something like:

COMPOSE_PROJECT_NAME=nerf_optimization_e23zhou
FIXUID=1011
FIXGID=1014

Inside Instant NGP Container

cmake -DNGP_BUILD_WITH_GUI=off ./ -B ./build
cmake --build build --config RelWithDebInfo -j 16

to build and run the thing, do not use the instant-ngp executable, instead use

python3 scripts/run.py path/to/data_images

Inside Kilonerf Container

BEFORE YOU ENTER make sure that you mount a your downloaded nsvf dataset

wget https://dl.fbaipublicfiles.com/nsvf/dataset/Synthetic_NSVF.zip && unzip -n Synthetic_NSVF.zip
wget https://dl.fbaipublicfiles.com/nsvf/dataset/Synthetic_NeRF.zip && unzip -n Synthetic_NeRF.zip
wget https://dl.fbaipublicfiles.com/nsvf/dataset/BlendedMVS.zip && unzip -n BlendedMVS.zip
wget https://dl.fbaipublicfiles.com/nsvf/dataset/TanksAndTemple.zip && unzip -n TanksAndTemple.zip

Inside compose, mount to /home/docker/kilonerf/data/nsvf

Inside the container, compile KiloNeRF's C++/CUDA code

cd $KILONERF_HOME/cuda
python setup.py develop

To benchmark a trained model run:
bash benchmark.sh

You can launch the interactive viewer by running:
bash render_to_screen.sh

To train a model yourself run
bash train.sh

The default dataset is Synthetic_NeRF_Lego, you can adjust the dataset by setting the dataset variable in the respective script.

Nerf PyTorch

You can follow the quickstart while inside the container

Download data for two example datasets: lego and fern

bash download_example_data.sh

To train a low-res lego NeRF:

python run_nerf.py --config configs/lego.txt

After training for 100k iterations (~4 hours on a single 2080 Ti), you can find the following video at logs/lego_test/lego_test_spiral_100000_rgb.mp4.

To copy things out

scp e23zhou@guacamole:/home/e23zhou/code/nerf-optimization/src/nerf-pytorch/logs/blender_paper_lego/blender_paper_lego_spiral_200000_rgb.mp4 /home/edwardius/code