Skip to content

SaigeModelInfo

  • #include <SaigeVisionAPI.h>

Structure containing model information.

Public Attributes

Type Name
char[ ] networkType
Indicates the type of deep learning network used.
int numOfClasses
Number of classes.
char[ ][ ] classNames
char[ ][ ] classColors
Class color codes. Colors are in HTML format and can be converted to RGB values using the SaigeColorCodeToRGB API.
bool[ ] classIsNg
Indicates whether each class is an NG class.
const char * description
Description provided for the model at training time.
const char * usedGpuName
Name of the GPU used for training.
const char * numOfIterations
Number of training iterations completed.
int numOfTrainImages
Number of images used directly in training.
int numOfValidationImages
Number of images used for validation during training.
long long trainingTimeStarted
Unix timestamp representing the start time of training.
long long trainingTimeCompleted
Unix timestamp representing the completion time of training.
long long trainingTimeElapsed
Time taken to complete training, in seconds.
char[ ] developerVersion
Version of SaigeVision Developer used during model training.
char[ ] serverVersion
Server version of the SaigeVision Developer used during model training.
char[ ] moduleVersion
Deep learning module version of the SaigeVision Developer used during model training.
SaigeRoiInfo roiInfo
ROI information.
float resizeFactor
(SEG) Resize factor applied to the input image, also applied to the inference image.
int cropSizeWidth
(DET, SEG, IEN) Width of the patch used during training, unrelated to inference images.
int cropSizeHeight
(DET, SEG, IEN) Height of the patch used during training, unrelated to inference images.
int imageResizeWidth
(IAD, CLS, DET) Width for resizing input images, applied to inference images as well.
int imageResizeHeight
(IAD, CLS, DET) Height for resizing input images, applied to inference images as well.
int targetTextSize
(OCR) Target text box height used in training, automatically adjusts the resize scale for the text box in the image.
const char * targetImageQuality
(IEN) Selected enhancement quality during training, either high or ultra-high.
const char * inputDataProcessingMode
(DET) Data preprocessing mode used during training.
int cropNumOfRows
(DET) Number of rows in the cropped image when the data preprocessing mode is crop_and_resize.
int cropNumOfCols
(DET) Number of columns in the cropped image when the data preprocessing mode is crop_and_resize.
const char * targetTextShape
(OCR) Target text shape used during training, either straight or curved. This may be empty for models trained in older versions.

Public Attributes Documentation

networkType

Indicates the type of deep learning network used. 

char[] networkType;

Network types ending with (v1) refer to network types used in SaigeVision 1.x.x versions (although the network type is the same, model files are not compatible).

Functionality Supported Network Types
IAD Fast, Standard, Accurate
CLS Fast, Standard, Accurate, Shallow (V1), Standard (V1), Deep (V1)
SEG Fast, Standard, Accurate, Shallow (V1), Standard (V1), Deep (V1)
DET Fast, Standard, Accurate, Extended
OCR Combination of Fast, Standard / Contextual, Literal / Short, Medium,

Long (e.g., Industry, Fast Short) For more details, refer to the SAIGE VISION Developer documentation. IEN | Fast, Standard

numOfClasses

Number of classes. 

int numOfClasses;

The number of classes is calculated as follows based on functionality:

Functionality Number of Classes Note
IAD 2 Fixed as OK and NG
CLS Number of user-added classes -
SEG Number of user-added classes + 1 Includes automatically allocated OK class
DET Number of user-added classes + 1 Includes automatically allocated OK class
OCR - Not applicable
IEN - Not applicable

classNames

char[][] classNames;

class names

classColors

Class color codes. Colors are in HTML format and can be converted to RGB values using the SaigeColorCodeToRGB API. 

char[][] classColors;

See also: SaigeColorCodeToRGB

classIsNg

Indicates whether each class is an NG class. 

bool[] classIsNg;

description

Description provided for the model at training time. 

const char * description;

usedGpuName

Name of the GPU used for training. 

const char * usedGpuName;

numOfIterations

Number of training iterations completed. 

const char * numOfIterations;

numOfTrainImages

Number of images used directly in training. 

int numOfTrainImages;

numOfValidationImages

Number of images used for validation during training. 

int numOfValidationImages;

trainingTimeStarted

Unix timestamp representing the start time of training. 

long long trainingTimeStarted;

trainingTimeCompleted

Unix timestamp representing the completion time of training. 

long long trainingTimeCompleted;

trainingTimeElapsed

Time taken to complete training, in seconds. 

long long trainingTimeElapsed;

developerVersion

Version of SaigeVision Developer used during model training. 

char[] developerVersion;

serverVersion

Server version of the SaigeVision Developer used during model training. 

char[] serverVersion;

moduleVersion

Deep learning module version of the SaigeVision Developer used during model training. 

char[] moduleVersion;

roiInfo

ROI information. 

SaigeRoiInfo roiInfo;

See also: SaigeRoiInfo

resizeFactor

(SEG) Resize factor applied to the input image, also applied to the inference image. 

float resizeFactor;

cropSizeWidth

(DET, SEG, IEN) Width of the patch used during training, unrelated to inference images. 

int cropSizeWidth;

cropSizeHeight

(DET, SEG, IEN) Height of the patch used during training, unrelated to inference images. 

int cropSizeHeight;

imageResizeWidth

(IAD, CLS, DET) Width for resizing input images, applied to inference images as well. 

int imageResizeWidth;

imageResizeHeight

(IAD, CLS, DET) Height for resizing input images, applied to inference images as well. 

int imageResizeHeight;

targetTextSize

(OCR) Target text box height used in training, automatically adjusts the resize scale for the text box in the image. 

int targetTextSize;

targetImageQuality

(IEN) Selected enhancement quality during training, either high or ultra-high. 

const char * targetImageQuality;

inputDataProcessingMode

(DET) Data preprocessing mode used during training. 

const char * inputDataProcessingMode;

Either resize or crop_and_resize. For models trained in older versions, it may be resize or crop.

cropNumOfRows

(DET) Number of rows in the cropped image when the data preprocessing mode is crop_and_resize. 

int cropNumOfRows;

cropNumOfCols

(DET) Number of columns in the cropped image when the data preprocessing mode is crop_and_resize. 

int cropNumOfCols;

targetTextShape

(OCR) Target text shape used during training, either straight or curved. This may be empty for models trained in older versions. 

const char * targetTextShape;