DDNeRF: Depth Distribution Neural Radiance Fields

This is the official repo for the paper: Depth Distribution Neural Radiance Fields. This repo also contains torch implementation of MipNeRF.

See project page for more details - https://dadonda89.github.io/ddnerfweb/

Some functions in this repo were taken from:

- https://github.com/Fyusion/LLFF

- https://github.com/yenchenlin/nerf-pytorch

- https://github.com/google/mipnerf

Create and activate conda environment:

conda env create -f environment.yml
conda activate nerf

Train model:

 python train_model.py --config configs/XXXX.yaml

Eval model:

python eval_nerf.py --logdir path_to_model_logdir

Video rendering:

python render_video.py --logdir path_to_model_logdir    

###example configs:

1) config_360.yml - for real world 360 bounded scene using ddnerf model

2) config_ff.yml - for real world forward facing scenes using ddnerf model

3) config_blender.yml - for synthetic scenes using ddnerf model

4) config_XXX_mipnerf.yml - for XXX scenes using mipnerf model

Data:

1) for real world data - the code support data format after colmap sparse reconstruction.

2) synthetic data format is similar to NeRF blender dataset

3) example real world data (motorbike scene) can be downloaded from: TBD