benchmark

Running a benchmark

run_benchmark.py runs the cone flattening code on a dataset that you provide and records the algorithm's performance. The script produces a tsv file recording some performance statistics for each mesh in the input. These files can then be analyed by summarize_results.py.

flag	purpose
--dataset_dir=/path/to/meshes	Directory of mesh files to run on (required).
--output_dir=/path/to/output/dir	Directory to store results (required).
--good_list=meshes_to_use.txt	File of meshes which should be used. By default, all meshes are used.
--bad_list=meshes_to_skip.txt	File of meshes which should be excluded.
--n_threads=1	Number of threads to run on (default=1).
--timeout=600	Timeout in seconds (default=600).
--max_meshes=1000	Maximum number of meshes to process. By default, all meshes are used.
--use_ffield_cones	Use cones from an MPZ-style .ffield file (must have the same base name as the mesh file, and be in the same directory.
--save_parameterized_meshes	Save parameterizations as well as performance statistics.

Summarizing benchmark results

Run process_results.py your_output_dir to summarize the results and plot the algorithm runtime (saved to your_output_dir/analysis) along with a csv containing the statistics from all meshes.

flag	purpose
--name=DatasetName	Dataset name to use in figure captions.
--plot_runtime	Plot the algorithm runtime.
--merged_files	Name of a csv to read, instead of reading in individual mesh records.