#include "darknet.h"

Collaboration diagram for network:

[legend]

Public Attributes
int	adam

int	adversarial

float	adversarial_lr

float	angle

float	aspect

int	attention

int	augment_speed

float	B1

float	B2

float *	badlabels_reject_threshold

float	badlabels_rejection_percentage

int	batch

int	batches_cycle_mult

int	batches_per_cycle

int	benchmark_layers

int	blur

int	burn_in

int	c
	The number of channels for the network. Typically `3` when working with RGB images.

int	center

float	clip

int	contrastive

int	contrastive_color

int	contrastive_jit_flip

float *	cost

void *	cuda_graph

void *	cuda_graph_exec

int *	cuda_graph_ready

int	cudnn_half

int *	cur_iteration

int	current_subdivision

float	decay

float *	delta

float *	delta_gpu

float *	delta_rolling_avg

float *	delta_rolling_max

float *	delta_rolling_std

Darknet::NetworkDetails *	details

int	dynamic_minibatch

float	ema_alpha

float	epoch

float	eps

int	equidistant_point

float	exposure

int	flip
	horizontal flip 50% probability augmentaiont for classifier training (default = 1)

float	gamma

int	gaussian_noise

float *	global_delta_gpu

int	gpu_index

int	h
	The height of the network. Must be divisible by `32`. E.g, `480`.

tree *	hierarchy

float	hue

int	index

int	init_sequential_subdivisions

float *	input

float **	input16_gpu

float **	input_gpu

float *	input_pinned_cpu
	memory allocated using `cudaHostAlloc()` which is used to transfer between the GPU and CPU

int	input_pinned_cpu_flag

float *	input_state_gpu

int	inputs

float	label_smooth_eps

Darknet::Layer *	layers

float	learning_rate

float	learning_rate_max

float	learning_rate_min

int	letter_box

float	loss_scale

int	max_batches

float	max_chart_loss

int	max_crop

size_t	max_delta_gpu_size

size_t *	max_input16_size

size_t *	max_output16_size

float	max_ratio

int	min_crop

float	min_ratio

int	mixup

float	momentum

int	mosaic_bound

int	n
	the number of layers in the network

int	notruth

int	num_boxes

float	num_sigmas_reject_badlabels

int	num_steps

int	optimized_memory

float *	output

float **	output16_gpu

float *	output_gpu

int	outputs

learning_rate_policy	policy

float	power

int	random

int	resize_step

int *	rewritten_bbox

float	saturation

float	scale

float *	scales

uint64_t *	seen

float *	seq_scales

int	sequential_subdivisions

float *	state_delta_gpu

int	step

int *	steps

int	subdivisions

int *	t

int	time_steps

int *	total_bbox

int	track

int	train

int	train_images_num

float *	truth

float **	truth_gpu

int	truths

int	try_fix_nan

int	unsupervised

int	use_cuda_graph

int	w
	The width of the network. Must be divisible by `32`. E.g., `640`.

int	wait_stream

int	weights_reject_freq

float *	workspace

size_t	workspace_size_limit

Member Data Documentation

◆ adam

int network::adam

◆ adversarial

int network::adversarial

◆ adversarial_lr

float network::adversarial_lr

◆ angle

float network::angle

◆ aspect

float network::aspect

◆ attention

int network::attention

◆ augment_speed

int network::augment_speed

◆ B1

float network::B1

◆ B2

float network::B2

◆ badlabels_reject_threshold

float* network::badlabels_reject_threshold

◆ badlabels_rejection_percentage

float network::badlabels_rejection_percentage

◆ batch

int network::batch

◆ batches_cycle_mult

int network::batches_cycle_mult

◆ batches_per_cycle

int network::batches_per_cycle

◆ benchmark_layers

int network::benchmark_layers

◆ blur

int network::blur

◆ burn_in

int network::burn_in

◆ c

int network::c

The number of channels for the network. Typically 3 when working with RGB images.

◆ center

int network::center

◆ clip

float network::clip

◆ contrastive

int network::contrastive

◆ contrastive_color

int network::contrastive_color

◆ contrastive_jit_flip

int network::contrastive_jit_flip

◆ cost

float* network::cost

◆ cuda_graph

void* network::cuda_graph

◆ cuda_graph_exec

void* network::cuda_graph_exec

◆ cuda_graph_ready

int* network::cuda_graph_ready

◆ cudnn_half

int network::cudnn_half

◆ cur_iteration

int* network::cur_iteration

◆ current_subdivision

int network::current_subdivision

◆ decay

float network::decay

◆ delta

float* network::delta

◆ delta_gpu

float* network::delta_gpu

◆ delta_rolling_avg

float* network::delta_rolling_avg

◆ delta_rolling_max

float* network::delta_rolling_max

◆ delta_rolling_std

float* network::delta_rolling_std

◆ details

Darknet::NetworkDetails* network::details

◆ dynamic_minibatch

int network::dynamic_minibatch

◆ ema_alpha

float network::ema_alpha

◆ epoch

float network::epoch

◆ eps

float network::eps

◆ equidistant_point

int network::equidistant_point

◆ exposure

float network::exposure

◆ flip

int network::flip

horizontal flip 50% probability augmentaiont for classifier training (default = 1)

◆ gamma

float network::gamma

◆ gaussian_noise

int network::gaussian_noise

◆ global_delta_gpu

float* network::global_delta_gpu

◆ gpu_index

int network::gpu_index

◆ h

int network::h

The height of the network. Must be divisible by 32. E.g, 480.

◆ hierarchy

tree* network::hierarchy

◆ hue

float network::hue

◆ index

int network::index

◆ init_sequential_subdivisions

int network::init_sequential_subdivisions

◆ input

float* network::input

◆ input16_gpu

float** network::input16_gpu

◆ input_gpu

float** network::input_gpu

◆ input_pinned_cpu

float* network::input_pinned_cpu

memory allocated using cudaHostAlloc() which is used to transfer between the GPU and CPU

◆ input_pinned_cpu_flag

int network::input_pinned_cpu_flag

◆ input_state_gpu

float* network::input_state_gpu

◆ inputs

int network::inputs

◆ label_smooth_eps

float network::label_smooth_eps

◆ layers

Darknet::Layer* network::layers

◆ learning_rate

float network::learning_rate

◆ learning_rate_max

float network::learning_rate_max

◆ learning_rate_min

float network::learning_rate_min

◆ letter_box

int network::letter_box

◆ loss_scale

float network::loss_scale

◆ max_batches

int network::max_batches

◆ max_chart_loss

float network::max_chart_loss

◆ max_crop

int network::max_crop

◆ max_delta_gpu_size

size_t network::max_delta_gpu_size

◆ max_input16_size

size_t* network::max_input16_size

◆ max_output16_size

size_t* network::max_output16_size

◆ max_ratio

float network::max_ratio

◆ min_crop

int network::min_crop

◆ min_ratio

float network::min_ratio

◆ mixup

int network::mixup

◆ momentum

float network::momentum

◆ mosaic_bound

int network::mosaic_bound

◆ n

int network::n

the number of layers in the network

◆ notruth

int network::notruth

◆ num_boxes

int network::num_boxes

◆ num_sigmas_reject_badlabels

float network::num_sigmas_reject_badlabels

◆ num_steps

int network::num_steps

◆ optimized_memory

int network::optimized_memory

◆ output

float* network::output

◆ output16_gpu

float** network::output16_gpu

◆ output_gpu

float* network::output_gpu

◆ outputs

int network::outputs

◆ policy

learning_rate_policy network::policy

◆ power

float network::power

◆ random

int network::random

◆ resize_step

int network::resize_step

◆ rewritten_bbox

int* network::rewritten_bbox

◆ saturation

float network::saturation

◆ scale

float network::scale

◆ scales

float* network::scales

◆ seen

uint64_t* network::seen

◆ seq_scales

float* network::seq_scales

◆ sequential_subdivisions

int network::sequential_subdivisions

◆ state_delta_gpu

float* network::state_delta_gpu

◆ step

int network::step

◆ steps

int* network::steps

◆ subdivisions

int network::subdivisions

◆ t

int* network::t

◆ time_steps

int network::time_steps

◆ total_bbox

int* network::total_bbox

◆ track

int network::track

◆ train

int network::train

◆ train_images_num

int network::train_images_num

◆ truth

float* network::truth

◆ truth_gpu

float** network::truth_gpu

◆ truths

int network::truths

◆ try_fix_nan

int network::try_fix_nan

◆ unsupervised

int network::unsupervised

◆ use_cuda_graph

int network::use_cuda_graph

◆ w

int network::w

The width of the network. Must be divisible by 32. E.g., 640.

◆ wait_stream

int network::wait_stream

◆ weights_reject_freq

int network::weights_reject_freq

◆ workspace

float* network::workspace

◆ workspace_size_limit

size_t network::workspace_size_limit

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

darknet.h

Public Attributes