git clone https://github.com/ashawkey/nerf_template.git
cd nerf_template

pip install -r requirements.txt

# install all extension modules
bash scripts/install_ext.sh

# if you want to install manually, here is an example:
cd raymarching
python setup.py build_ext --inplace # build ext only, do not install (only can be used in the parent directory)
pip install . # install to python path (you still need the raymarching/ folder, since this only install the built extension.)

# prepare your video or images under /data/custom, and run colmap (assumed installed):
python scripts/colmap2nerf.py --video ./data/custom/video.mp4 --run_colmap # if use video
python scripts/colmap2nerf.py --images ./data/custom/images/ --run_colmap # if use images

## -O: instant-ngp
# prune sampling points by maintaining a density grid
python main.py data/bonsai/ --workspace trial_bonsai_ngp --enable_cam_center --downscale 4 -O --background random --bound 8

## -O2: nerfstudio nerfacto
# use proposal network to predict sampling points
python main.py data/bonsai/ --workspace trial_bonsai_nerfacto --enable_cam_center --downscale 4 -O2

# MeRF network backbone
python main.py data/bonsai/ --workspace trial_bonsai_nerfacto --enable_cam_center --downscale 4 -O2 --backbone merf

### -O: equals
--fp16 --preload
--cuda_ray --mark_untrained
--adaptive_num_rays --random_image_batch

### -O2: equals
--fp16 --preload
--contract --bound 128
--adaptive_num_rays --random_image_batch 

### load checkpoint
--ckpt latest # by default we load the latest checkpoint in the workspace
--ckpt scratch # train from scratch.
--ckpt trial/checkpoints/xxx.pth # specify it by path

### training
--num_rays 4096 # number of rays to evaluate per training step
--adaptive_num_rays # ignore --num_rays and use --num_points to dynamically adjust number of rays.
--num_points 262144 # targeted number of points to evaluate per training step (to adjust num_rays)

### testing
--test # test, save video and mesh
--test_no_video # do not save video
--test_no_mesh # do not save mesh

### dataset related
--data_format [colmap|nerf] # dataset format
--enable_cam_center # use camera center instead of sparse point cloud center as the scene center (colmap dataset only) (only for 360-degree captured datasets, do not use this for forward-facing datasets!)
--enable_cam_near_far # estimate camera near & far from sparse points (colmap dataset only)

--bound 16 # scene bound set to [-16, 16]^3.
--scale 0.3 # camera scale, if not specified, automatically estimate one based on camera positions.

### visualization 
--vis_pose # viusalize camera poses and sparse points (sparse points are colmap dataset only)
--gui # open gui (only for testing, training in gui is not well supported!)

### balance between surface quality / rendering quality

# increase these weights to get better surface quality but worse rendering quality
--lambda_tv 1e-7 # total variation loss
--lambda_entropy 1e-3 # entropy on rendering weights (transparency, alpha), encourage them to be either 0 or 1