#include <ml/neural_net/image_augmentation.hpp>

Public Attributes
size_t	batch_size = 0

size_t	output_width = 0

size_t	output_height = 0

int	random_seed = 0

float	crop_prob = 0.f

float	pad_prob = 0.f

float	horizontal_flip_prob = 0.f

float	brightness_max_jitter = 0.f

float	contrast_max_jitter = 0.f

float	saturation_max_jitter = 0.f

float	hue_max_jitter = 0.f

Detailed Description

Parameters for constructing new image_augmenter instances.

Default constructed values perform no augmentation, outside of resizing to the output width and height (which must be specified).

Definition at line 165 of file image_augmentation.hpp.

Member Data Documentation

◆ batch_size

size_t turi::neural_net::image_augmenter::options::batch_size = 0

The N dimension of the resulting float array.

Definition at line 168 of file image_augmentation.hpp.

◆ brightness_max_jitter

float turi::neural_net::image_augmenter::options::brightness_max_jitter = 0.f

Maximum pixel value to add or subtract to each channel.

For example, a value of 0.05 adds a random value between -0.05 and 0.05 to each channel of each pixel (represented as a value from 0 to 1).

Definition at line 200 of file image_augmentation.hpp.

◆ contrast_max_jitter

float turi::neural_net::image_augmenter::options::contrast_max_jitter = 0.f

Maximum proportion to increase or decrease contrast.

For example, a value of 0.05 multiplies the contrast by a random value between 0.95 and 1.05.

Definition at line 208 of file image_augmentation.hpp.

◆ crop_prob

float turi::neural_net::image_augmenter::options::crop_prob = 0.f

The probability of applying (attempting) a random crop.

Definition at line 180 of file image_augmentation.hpp.

◆ horizontal_flip_prob

float turi::neural_net::image_augmenter::options::horizontal_flip_prob = 0.f

The probability of flipping the image horizontally.

Definition at line 188 of file image_augmentation.hpp.

◆ hue_max_jitter

float turi::neural_net::image_augmenter::options::hue_max_jitter = 0.f

Maximum proportion to rotate the hues.

For example, a value of 0.05 applies a random rotation between -0.05 * pi and 0.05 * pi.

Definition at line 224 of file image_augmentation.hpp.

◆ output_height

size_t turi::neural_net::image_augmenter::options::output_height = 0

The H dimension of the resulting float array.

Definition at line 174 of file image_augmentation.hpp.

◆ output_width

size_t turi::neural_net::image_augmenter::options::output_width = 0

The W dimension of the resulting float array.

Definition at line 171 of file image_augmentation.hpp.

◆ pad_prob

float turi::neural_net::image_augmenter::options::pad_prob = 0.f

The probability of applying (attempting) a random pad.

Definition at line 184 of file image_augmentation.hpp.

◆ random_seed

int turi::neural_net::image_augmenter::options::random_seed = 0

Seed for all pseudo-random number generation used by augmentation.

Definition at line 177 of file image_augmentation.hpp.

◆ saturation_max_jitter

float turi::neural_net::image_augmenter::options::saturation_max_jitter = 0.f

Maximum proportion to increase or decrease saturation.

For example, a value of 0.05 multiplies the saturation by a random value between 0.95 and 1.05.

Definition at line 216 of file image_augmentation.hpp.

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

ml/neural_net/image_augmentation.hpp

Public Attributes