smvs cli options - natowi/meshroom_external_plugins GitHub Wiki

Shading aware Multi-View Stereo https://github.com/flanggut/smvs

Usage: smvsrecon [ OPTS ] SCENE_DIR
Available options:
  -a, --alpha=ARG          Regularization parameter, a higher alpha leads to smoother surfaces [1]
  -s, --scale=ARG          Scale of input images [estimated to reduce images to a maximum of ~1.5MP]
  -i, --image=ARG          Image embedding [undistorted]
  -n, --neighbors=ARG      Number of neighbors for recon [6]
  -o, --output-scale=ARG   Scale of output depth [2]
  -l, --list-view=ARG      Reconstructs given viewIDs (given as string "0-10")
  -t, --threads=ARG        Number of threads [Num CPU cores]. Peak memory requirement is ~1GB per thread and 2 megapixel image resolution
  -d, --debug-lvl=ARG      Debug level [0]
  -r, --recon-only         Generate only depth maps and no output ply. [off]
  -M, --max-pixels=ARG     Maximal number of pixels for reconstruction. Images will be rescaled to be below this value. [1700000]
  -S, --shading            Use shading-based optimization. [off]
  -R, --regularize-lighting=ARG  Use additional basic surface regularization when lighting is turned on. This is untested but may improve results for scenes with complex lighting (0 = off, 100 = use full basic regularizer). [0]
  -g, --gamma-srgb         Apply inverse SRGB gamma correction. [off]
  -m, --mesh               Create Triangle mesh instead of simple point cloud (WIP). [off]
  -y, --simplify           Create simplified triangle mesh (WIP). [off]
  -f, --force              Force reconstruction of result embeddings
  --no-cut                 Turn off surface cutting and export fill pointcloud from all depth values. [on]
  --aabb=ARG               Comma separated AABB for output: min,min,min,max,max,max
  --min-neighbors=ARG      Minimal number of neighbors for reconstruction. [3]
  --no-sgm                 Turn off semi-global matching.
  --force-sgm              Force reconstruction of SGM embeddings.
  --sgm-scale=ARG          Scale of reconstruction of SGM embeddings relative to input scale. [1]
  --sgm-range=ARG          Range for SGM depth sweep, given as string "0.1,3.5". [estimated from SfM pointcloud] (this option is untested please report any issues)
  --full-opt               Run full optimization on all nodes (otherwise it only runs on non converged nodes) [off]
  --clean                  Clean scene and remove all result embeddings