- binary spherical gradient pattern reconstruction
- generating ply file
- normal, poisition combination
-
you need to install suitesparse library to run.
-
edit or create .fc file to run. .fc file contains (fx, fy, width/2(center_x), height/2(center_y)). Take notice of sign. The sign and magnitude (axis) should match with those of the normals we obtained using light stage.
-
build :
make
- simply run :
bash run.sh
-
You may change some configurations (recon.conf) such as focal length, sensor size.
-
binary_reconstruction.py :
python binary_reconstruction.py [-V] [-format "format of input files"] [-path "path of input files"]
- generate_pointcloud :
python generate_pointcloud_ns.py [color file path] [depth map path] [normal map path] [output file path]
- mesh_opt : follow instruction in link. You can use .fc files (range grid) or you can just run mesh using trinangulation. As documented on the link above, using it generates much qualified results.
./mesh_opt result/before.ply -fc emily.fc -lambda 0.01 norm:result/fixed.ply
./mesh_opt result/before.ply -lambda 0.01 norm:result/fixed.ply
./mesh_opt result/before.ply -lambda 0.01 result/fixed.ply
- you may change "-lambda" value change weights for optimization. Also, you may get rid of "norm" flag to save only point cloud data, no normals.