Forked repo of "Official PyTorch implementation of SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering"
to remove background contribution in the mesh, I rescale bg_bbox_factor from 4 to 1 (as for angel and statues), in extract_mesh_from_coarse_sugar() in sugar_extractors/coarse_mesh.py
1. Training using train.py (calls the scrips automatically)
1.1 Align cameras using COLMAP implementation of GS
python gaussian_splatting/convert.py -s data/Bureau/
1.2 Train GS model
python gaussian_splatting/train.py -s data/Bureau_sdf_low_poly/ --iterations 7000 -m outputs_GS/Bureau
Train the Sugar model
python train.py -s data/Bureau_sdf_low_poly/ -c outputs_GS/Bureau/ -r "sdf" --low_poly True
2. Running separate scripts
2.1 extract_mesh.py
bg_bbox_factor
from 4 to 1 (as for angel and statues), in extract_mesh_from_coarse_sugar() in sugar_extractors/coarse_mesh.pybg_bbox_factor
is only present in the extract_mesh.py2.2 train_refined.py
2.3 extract_refined_mesh_with_texture.py
python extract_refined_mesh_with_texture.py -s data/Bureau_sdf_low_poly_bgfactor_1/ -c outputs_GS/Bureau/ -i 7000 -m output/refined/Bureau_sdf_low_poly_bgfactor_1/sugarfine_3Dgs7000_sdfestim02_sdfnorm02_level03_decim200000_normalconsistency01_gaussperface1/15000.pt -o output/refined_mesh/Bureau_sdf_low_poly_bgfactor_1