Перейти до основного вмісту

NormalizationOptions Struct

UInt8 RGB/BGR/GRAY8 to dense BF16; coefficients are in semantic RGB order. More...

Declaration​

struct simaai::neat::NormalizationOptions { ... }

Included Headers​

#include <ImageTransform.h>

Public Member Attributes Index​

std::array< float, 3 >mean {0.0f, 0.0f, 0.0f}
std::array< float, 3 >stddev {1.0f, 1.0f, 1.0f}
std::stringelement_name
std::optional< float >rescale_factor

Explicit source rescale; omitted retains the legacy /255 affine construction. More...

PreprocFloatingConversionfloating_conversion = PreprocFloatingConversion::Legacy

Description​

UInt8 RGB/BGR/GRAY8 to dense BF16; coefficients are in semantic RGB order.

Without rescale_factor, arithmetic retains legacy Preproc coefficients: (x - float(mean * 255)) * float(1 / (stddev * 255)). An explicit source factor instead uses FP32 rescale, subtract and divide. No resize, color conversion, tessellation or implicit CPU fallback is performed.

Definition at line 62 of file ImageTransform.h.

Public Member Attributes​

element_name​

std::string simaai::neat::NormalizationOptions::element_name

Definition at line 65 of file ImageTransform.h.

floating_conversion​

PreprocFloatingConversion simaai::neat::NormalizationOptions::floating_conversion = PreprocFloatingConversion::Legacy

Definition at line 68 of file ImageTransform.h.

mean​

std::array<float, 3> simaai::neat::NormalizationOptions::mean {0.0f, 0.0f, 0.0f}

Definition at line 63 of file ImageTransform.h.

rescale_factor​

std::optional<float> simaai::neat::NormalizationOptions::rescale_factor

Explicit source rescale; omitted retains the legacy /255 affine construction.

Definition at line 67 of file ImageTransform.h.

stddev​

std::array<float, 3> simaai::neat::NormalizationOptions::stddev {1.0f, 1.0f, 1.0f}

Definition at line 64 of file ImageTransform.h.

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

Generated via doxygen2docusaurus 2.0.0 by Doxygen 1.15.0.