A Customized NeRF pipeline by Qi
To install the environment, you just need to type (assume anaconda or miniconda installed),
make setup_env
Or using scripts, after using conda (conda create -n nerf -y python=3.8, conda activate nerf) or it will on your default python environment.
bash ./scripts/env_setting.sh
Or using requirements.txt, python -m pip install -r ./requirements.txt.
To get a minimal dataset for traning demo, you just need to type
make download_data
Or using scripts,
bash ./scripts/download_data.sh
To get more dataset, see data/README.md for details.
Please refer to preprocess_data/README.md for more information to a customized dataset (currently it was not fully supported).
To train a lego for yourself after the environment had been set up, you just need to type
make train_lego
Or using scripts,
bash ./scripts/train_lego.sh
This will make a log directory for check points, visualizations, etc.
See makefile for other commands, and see arg_parser.py for more options on trainning.
You can download pretrained model from our release page, and place then in ./pretrained fold.
To extra a mesh yourself, you just need to type
make extra_mesh_demo
Or in extra_mesh.sh
bash ./srcipts/extra_mesh.sh
To be more customized, see extra_mesh/README.md for details.
Please see makefile for more information.