This is the official implementation of Normal-guided Detail-Preserving Neural Implicit Functions for High-Fidelity 3D Surface Reconstruction.

git clone https://github.com/sn-nir/sn-nir.git
cd sn-nir
pip install -r requirements.txt

Our data format is inspired from IDR as follows:

CASE_NAME
|-- cameras.npz    # camera parameters
|-- image
    |-- 000.png        # image for each view
    |-- 001.png
    ...
|-- normal
    |-- 000.png        # normal map for each view
    |-- 001.png
    ...
|-- depth
    |-- 000.png        # depth for each view
    |-- 001.png
    ...
|-- mask
    |-- 000.png        # mask for each view
    |-- 001.png
    ...

One can create folders with different data in it, for instance, a normal folder for each normal estimation method. The name of the folder must be set in the used .conf file.

Train

python run_experiments.py --mode train --conf ./confs/CONF_NAME.conf --case CASE_NAME

Extract mesh

python run_experiments.py --mode validate_mesh --conf ./confs/CONF_NAME.conf --case CASE_NAME --is_continue