This repo contains the photometric normal refinement routines.
For geometric coarse normal estimation, see this repo.
python>=3.7 numpy skimage matplotlib scipy python-opencv open3d==0.16.0
Implemented in stereo_flash_no_flash.py, evaluate.py, pointcloud_processing.py.
Blender simulation and exporting scripts in blender/.
Configuration is in configs.py.
Helper functions are in utils.py, cp_hw2.py.
Place the data in datasets/.
To perform reconstruction, and run the following steps:
Set the object name in configs.py. Currently supported: bunny, buddha, shoe.
Turn point cloud from SfM toolbox into a mesh and then into a coarse normal map.
python3 pointcloud_processing.pypython3 evaluate.py