Ground-truth accuracy report for a reconstructed point cloud. More...

#include <accuracy_metrics.hpp>

Public Member Functions
std::string	to_json (int indent=2) const

Public Attributes
std::size_t	cloud_points = 0
	finite reconstruction points scored
std::size_t	gt_points = 0
	finite (post-downsample) GT points scored
double	accuracy_mean = 0.0
	mean recon->GT distance [m]
double	accuracy_median = 0.0
	median recon->GT distance [m]
double	completeness_mean = 0.0
	mean GT->recon distance [m]
double	completeness_median = 0.0
	median GT->recon distance [m]
double	chamfer = 0.0
	(accuracy_mean + completeness_mean) / 2 [m]
double	precision = 0.0
	fraction of recon points within threshold
double	recall = 0.0
	fraction of GT points within threshold
double	fscore = 0.0
	2PR/(P+R)
float	threshold = 0.05f
	inlier threshold used [m]

Detailed Description

Ground-truth accuracy report for a reconstructed point cloud.

Standard reconstruction-benchmark definitions, all distances in meters:

accuracy: for each reconstruction point, distance to the nearest ground-truth point. Measures how far the reconstruction strays from the reference surface (lower is better).
completeness: for each ground-truth point, distance to the nearest reconstruction point. Measures how much of the reference surface was actually reconstructed (lower is better).
chamfer: mean of accuracy_mean and completeness_mean.
precision: fraction of reconstruction points within threshold of the ground truth.
recall: fraction of ground-truth points within threshold of the reconstruction.
fscore: harmonic mean 2*P*R/(P+R) (0 when P+R == 0).

Unlike the ground-truth-free QualityReport, this requires an external reference cloud (e.g. a Faro survey), and both clouds must share a common coordinate frame.

Definition at line 41 of file accuracy_metrics.hpp.

Member Function Documentation

◆ to_json()

std::string reusex::geometry::AccuracyReport::to_json ( int indent = 2 ) const

nodiscard

Member Data Documentation

◆ accuracy_mean

double reusex::geometry::AccuracyReport::accuracy_mean = 0.0

mean recon->GT distance [m]

Definition at line 45 of file accuracy_metrics.hpp.

◆ accuracy_median

double reusex::geometry::AccuracyReport::accuracy_median = 0.0

median recon->GT distance [m]

Definition at line 46 of file accuracy_metrics.hpp.

◆ chamfer

double reusex::geometry::AccuracyReport::chamfer = 0.0

(accuracy_mean + completeness_mean) / 2 [m]

Definition at line 51 of file accuracy_metrics.hpp.

◆ cloud_points

std::size_t reusex::geometry::AccuracyReport::cloud_points = 0

finite reconstruction points scored

Definition at line 42 of file accuracy_metrics.hpp.

◆ completeness_mean

double reusex::geometry::AccuracyReport::completeness_mean = 0.0

mean GT->recon distance [m]

Definition at line 48 of file accuracy_metrics.hpp.

◆ completeness_median

double reusex::geometry::AccuracyReport::completeness_median = 0.0

median GT->recon distance [m]

Definition at line 49 of file accuracy_metrics.hpp.

◆ fscore

double reusex::geometry::AccuracyReport::fscore = 0.0

2*P*R/(P+R)

Definition at line 55 of file accuracy_metrics.hpp.

◆ gt_points

std::size_t reusex::geometry::AccuracyReport::gt_points = 0

finite (post-downsample) GT points scored

Definition at line 43 of file accuracy_metrics.hpp.

◆ precision

double reusex::geometry::AccuracyReport::precision = 0.0

fraction of recon points within threshold

Definition at line 53 of file accuracy_metrics.hpp.

◆ recall

double reusex::geometry::AccuracyReport::recall = 0.0

fraction of GT points within threshold

Definition at line 54 of file accuracy_metrics.hpp.

◆ threshold

float reusex::geometry::AccuracyReport::threshold = 0.05f

inlier threshold used [m]

Definition at line 57 of file accuracy_metrics.hpp.

The documentation for this struct was generated from the following file:

libs/reusex/include/reconstruction/accuracy_metrics.hpp

Public Member Functions

Public Attributes