ImageSampler Class Reference

Class for selecting uncertain image samples for further analysis based on various confidence metrics. More...

#include <imagesampler.h>

Public Member Functions
	ImageSampler (const std::string &conf_path, int save_interval, const std::string &model_type)
	Constructor to initialize ImageSampler object with configuration file path. More...
int	sample (const void *raw_output, cv::Mat &img, bool save_sample)
	Selects uncertain image samples based on configured criteria. More...
float	margin_confidence (std::vector< float > &prob_dist, bool sorted)
	Calculates margin confidence (difference between top two probabilities) More...
float	least_confidence (std::vector< float > &prob_dist, bool sorted)
	Calculates least confidence (normalized maximum probability) More...
float	ratio_confidence (std::vector< float > &prob_dist, bool sorted)
	Calculates ratio confidence (ratio of top two probabilities) More...
float	entropy_confidence (std::vector< float > &prob_dist)
	Calculates entropy-based confidence. More...

Public Attributes
std::string	statSavepath
std::string	dataSavepath

Detailed Description

Class for selecting uncertain image samples for further analysis based on various confidence metrics.

Class for selecting uncertain image samples for further analysis.

Constructor & Destructor Documentation

ImageSampler()

ImageSampler::ImageSampler	(	const std::string &	conf_path,
		int	save_interval,
		const std::string &	model_type
	)

Constructor to initialize ImageSampler object with configuration file path.

Constructor to initialize ImageSampler object.

Parameters

configFilePath	Path to the configuration file
conf_path	Path to configuration file
saver	Saver object for saving sampling statistics

Member Function Documentation

entropy_confidence()

float ImageSampler::entropy_confidence

(

std::vector< float > &

prob_dist

)

Calculates entropy-based confidence.

Calculates entropy based confidence.

Parameters

probabilityDistribution

Vector of class probabilities

Returns

Entropy-based confidence score

Parameters

prob_dist	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Entropy confidence score

least_confidence()

float ImageSampler::least_confidence	(	std::vector< float > &	prob_dist,
		bool	sorted = false
	)

Calculates least confidence (normalized maximum probability)

Calculates least confidence.

Parameters

probabilityDistribution	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Least confidence score

Parameters

prob_dist	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

least confidence score

margin_confidence()

float ImageSampler::margin_confidence	(	std::vector< float > &	prob_dist,
		bool	sorted = false
	)

Calculates margin confidence (difference between top two probabilities)

Parameters

probabilityDistribution	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Margin confidence score

Parameters

prob_dist	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Margin confidence score

ratio_confidence()

float ImageSampler::ratio_confidence	(	std::vector< float > &	prob_dist,
		bool	sorted = false
	)

Calculates ratio confidence (ratio of top two probabilities)

Calculates ratio confidence (difference between top two probabilities)

Parameters

probabilityDistribution	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Ratio confidence score

Parameters

prob_dist	Vector of class probabilities
sorted	Flag indicating if probabilities are already sorted (default: false)

Returns

Margin confidence score

sample()

int ImageSampler::sample	(	const void *	raw_output,
		cv::Mat &	img,
		bool	save_sample
	)

Selects uncertain image samples based on configured criteria.

Parameters

classificationResults	Vector of confidence scores for each image prediction
image	OpenCV image matrix
saveSample	Flag indicating whether to save sampled images
results	Vector of confidence scores for each image prediction
uncertainty_sampling	Vector indicating uncertainty criteria for each sample
img	OpenCV image matrix
save_sample	Flag indicating whether to save sampled images

Returns

1 on success, error code on failure

---

The documentation for this class was generated from the following files:

• include/imagesampler.h
• src/sampling/imagesampler.cpp