Skip to main content

run_benchmark_subset_case

Function run_benchmark_subset_case 

Source
pub fn run_benchmark_subset_case<A: ReconstructionAlgorithm>(
    dataset_name: impl Into<String>,
    algorithm_configuration: impl Into<String>,
    algorithm: A,
    subset: &DatasetSubset,
) -> Result<(BenchmarkRecord, Option<ReconstructionResult>)>
Expand description

Runs a concrete algorithm on a resolved dataset subset and records its provenance.

Clones the subset into a validated problem and evaluates its optional cropped truth and valid-object mask. Records the resolved original frame/illumination identifiers and [row, column, height, width] detector crop even when the algorithm fails. The deterministic case identity includes those selections. Problem-construction errors are returned; reconstruction and metric failures remain in record.error, as in run_benchmark_case.

use fpm_rs::{algorithms::AlternatingProjection, benchmark::run_benchmark_subset_case,
    datasets::{DatasetLoader, FrameSelector}, Result};
let dataset = DatasetLoader::new("converted-fpm")?.load()?;
let subset = dataset.subset().frames(FrameSelector::EveryNth(2))
    .crop_pixels(0, 0, 16, 16)?.build()?;
let (record, result) = run_benchmark_subset_case(
    "converted-fpm", "iterations=10", AlternatingProjection::default().iterations(10),
    &subset,
)?;
assert_eq!(record.spatial_crop, Some([0, 0, 16, 16]));