Metrics calculated on counters. More...

#include <sb.h>

Data Fields
float	accuracy
	accuracy More...

float	recall
	recall More...

float	precision
	precision More...

float	f1_score
	f1-score More...

float	specificity
	specificity More...

Detailed Description

Metrics calculated on counters.

Depending on the context, counters can be referred to samples, blobs, pixels or images.

See also: sb_t_res; sb_project_detection; sb_project_get_res

Definition at line 12305 of file sb.h.

Field Documentation

◆ accuracy

float sb_t_metrics::accuracy

accuracy

In general equal to accuracy = (tp + tn) / (tp + tn + fp + fn)
In case of sb_t_stat::metrics it is equal to accuracy = (good + no_good) / (good + no_good + overkill + escape)

Definition at line 12312 of file sb.h.

◆ f1_score

float sb_t_metrics::f1_score

f1-score

In general equal to f1-score = 2*tp / (2*tp + fn + fp)
In case of sb_t_stat::metrics it is equal to f1-score = (2 * no_good) / (2 * no_good + overkill + escape).
It is also known as Sorensen-Dice coefficient (or more simply as Dice) and is widely used in pixel segmentation task to evaluate the similarity between ground truth and prediction.

Definition at line 12335 of file sb.h.

◆ precision

float sb_t_metrics::precision

precision

In general equal to precision = tp / (tp + fp)
In case of sb_t_stat::metrics it is equal to precision = no_good / (no_good + overkill) .

Definition at line 12326 of file sb.h.

◆ recall

float sb_t_metrics::recall

recall

In general equal to recall = tp / (tp + fn)
In case of sb_t_stat::metrics it is equal to recall = no_good / (no_good + escape)

Definition at line 12319 of file sb.h.

◆ specificity

float sb_t_metrics::specificity

specificity

In general equal to specificity = tn / (tn + fp)
In case of sb_t_stat::metrics it is equal to specificity = good / (good + overkill)

Definition at line 12342 of file sb.h.

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

sb.h