KwanBL/D-NeuS

Recovering Fine Details for Neural Implicit Surface Reconstruction (WACV2023)

D-NeuS

Recovering Fine Details for Neural Implicit Surface Reconstruction (WACV2023)

Code

D-NeuS is built on NeuS. It improves the surface reconstruction quality with fine details recovered.

Training

python exp_runner.py --mode train --case <e.g., scan24>

Extract surface from trained model

python exp_runner.py --mode validate_mesh --case <e.g., scan24> --is_continue

View interpolation

python exp_runner.py --mode interpolate_<img_idx_0>_<img_idx_1> --case <e.g., scan24> --is_continue 

Data

Download the pre-processed DTU data from this link, and put it under ./data folder.

Acknowledgement

The code is based on NeuS. Some code snippets and data pre-processing are borrowed from MVSDF. Thanks for these great works.