Skip to content

luczeng/multi-view-evaluation

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ETH3D Multi-View Evaluation Program

This tool is used for evaluating multi-view reconstruction methods in the ETH3D benchmark.

If you use this code for research, please cite our paper:

T. Schöps, J. L. Schönberger, S. Galliani, T. Sattler, K. Schindler, M. Pollefeys, A. Geiger, "A Multi-View Stereo Benchmark with High-Resolution Images and Multi-Camera Videos", Conference on Computer Vision and Pattern Recognition (CVPR), 2017. [Bibtex][PDF][Supplementary]

Example usage:

ETH3DMultiViewEvaluation --reconstruction_ply_path reconstruction.ply \
                         --ground_truth_mlp_path scan_alignment.mlp \
                         --tolerances 0.01,0.02,0.05,0.1,0.2,0.5

Description of required program arguments:

  • --reconstruction_ply_path: Path to the reconstructed point cloud (as PLY).
  • --ground_truth_mlp_path: Path to the MeshLab project file which defines the poses of the ground truth laser scan files.
  • --tolerances: Comma-separated list of tolerance values to evaluate with.

Description of optional program arguments:

  • --voxel_size (default 0.01): Size of voxels for normalising scores per volume.
  • --beam_start_radius_meters (default 0.5 * 0.00225): Size of beam at the laser scanner origin for free-space modeling.
  • --beam_divergence_halfangle_deg (default 0.011): Beam divergence for free-space modeling.
  • --completeness_cloud_output_path (default ""): If set to a path, completeness visualizations for each tolerance value are written to <path>.tolerance_<tolerance>.ply.
  • --accuracy_cloud_output_path (default ""): If set to a path, accuracy visualizations for each tolerance value are written to <path>.tolerance_<tolerance>.ply.

About

Multi-view evaluation program.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 98.0%
  • CMake 2.0%