CtrEditor/Documentation/paddleocr-api.md

PaddleOCRSharp

Base class for engine objects

Custom loading path for PaddleOCR.dll, default is empty. If specified, it needs to be assigned before engine instantiation.

Initialization

Get the current path of the program

Get the current path of the program

Convert Image to Byte[]

Release memory

Get underlying error information

Json helper class

Json deserialization

Model configuration object

det_infer model path

cls_infer model path

rec_infer model path

Full path of ppocr_keys.txt file

Table model configuration object

table_model_dir model path

Table recognition dictionary

OCR recognition parameters

Whether to use GPU; default false

GPU id, effective when using GPU; default 0

Requested GPU memory; default 4000

Number of threads for CPU prediction. When the machine has sufficient cores, the higher this value, the faster the prediction; default 10

Whether to use mkldnn library; default true

Whether to perform text detection; default true

Whether to perform text recognition; default true

Whether to perform text orientation classification; default false

When the input image length and width are greater than 960, the image is scaled proportionally so that the longest side is 960; default 960

Used to filter the binary image predicted by DB. Setting to 0.-0.3 has no significant effect on results; default 0.3

DB post-processing threshold for filtering boxes. If detection has missing boxes, this can be reduced accordingly; default 0.5

Represents the tightness of the text box. Smaller values mean the text box is closer to the text; default 1.6

Whether to use dilation on the output map, default false

true: use polygon box to calculate bbox; false: use rectangle box. Rectangle calculation is faster, polygon boxes are more accurate for curved text areas.

Whether to visualize the results. If true, the prediction results will be saved as an ocr_vis.png file in the current directory. Default false

Whether to use direction classifier, default false

Score threshold for direction classifier, default 0.9

Direction classifier batchsize, default 1

Recognition model batchsize, default 6

Recognition model input image height, default 48

Recognition model input image width, default 320

Whether to display prediction results, default false

When using GPU prediction, whether to enable tensorrt, default false

OCR modifiable parameters

Dynamically modify whether to detect. When OCRParameter.det=true, m_det can dynamically turn off the det parameter

Dynamically modify whether to recognize. When OCRParameter.rec=true, m_rec can dynamically turn off the rec parameter

When the input image length and width are greater than 960, the image is scaled proportionally so that the longest side is 960; default 960

Used to filter the binary image predicted by DB. Setting to 0.-0.3 has no significant effect on results; default 0.3. Effective when m_det=true

DB post-processing threshold for filtering boxes. If detection has missing boxes, this can be reduced accordingly; default 0.5. Effective when m_det=true

Represents the tightness of the text box. Smaller values mean the text box is closer to the text; default 1.6. Effective when m_det=true

OCR recognition result

List of text blocks

Recognition result text

Recognition result text in Json format

Return string format

Recognized text block

List of coordinate vertices around the text block

Text block content

Text recognition confidence

Angle classification confidence

Angle classification label

Return string format

Point object

X coordinate, in pixels

Y coordinate, in pixels

Default constructor

Constructor

Return string format

OCR structured recognition result

Table recognition result

Number of rows

Number of columns

List of cells

List of text blocks

Cell

Cell constructor

Row number

Column number

Text blocks

Recognized text

PaddleOCR text recognition engine object

Initialize OCR engine object with default parameters

Initialize OCR engine object with default parameters

Model configuration object, if null then default values are used

PaddleOCR recognition engine object initialization

Model configuration object, if null then default values are used Recognition parameters, if null then default values are used

PaddleOCR recognition engine object initialization

Model configuration object, if null then default values are used Recognition parameters in json string format

Get default configuration

Root directory

Perform text recognition on image file

Image file OCR recognition result

Perform text recognition on image object

Image OCR recognition result

Text recognition

Image memory stream OCR recognition result

Text recognition

Image base64 OCR recognition result

Text recognition

Image memory address Image width Image height Image channel, usually 3 or 1 OCR recognition result

Result parsing

Structured text recognition

Image Table recognition result

Calculate table splitting

Dynamically modify parameters after initialization

Modifiable parameter object Whether successful, calling before initialization will result in failure

Whether to enable rectangular processing of detection results. Single characters are easily detected as diamond shapes, processing them as rectangles can improve recognition accuracy. Only applicable for horizontal text.

Release object

PaddleOCR table recognition engine object

PaddleStructureEngine recognition engine object initialization

PaddleStructureEngine recognition engine object initialization

Model configuration object, if null then default values are used

PaddleStructureEngine recognition engine object initialization

Model configuration object, if null then default values are used Recognition parameters, if null then default values are used

PaddleStructureEngine recognition engine object initialization

Model configuration object, if null then default values are used Recognition parameters in Json format, if null then default values are used

Get default configuration

Root directory

Perform table text recognition on image file

Image file Table recognition result

Perform table text recognition on image object

Image Table recognition result

Perform table text recognition on image Byte array

Image byte array Table recognition result

Perform table text recognition on image Base64

Image Base64 Table recognition result

Result parsing

Release object

OCR recognition parameters

When the input image length and width are greater than 488, the image is scaled proportionally, default 488

Whether to merge empty cells

Batch recognition quantity