PredictModel

BinaryToText is optional.

Specify either the Content field or the Body field. If you use an image URL, set Content to the image access address. If you use Base64 encoding content, set the Body field and set BinaryToText to true.

PDF files are limited to 20 MB and 10 pages. Except for long document model prediction, other prediction services only process the first page.

Description of the data fields returned by the long document information extraction model:

originalFileUrl     The URL of the original file.
predictFile         The collection of image URLs parsed for prediction.
data                The specific prediction results.
angle               The angle of the image. This field is returned only when NeedRotate is set to true. 0 indicates upright, 90 indicates the image faces right, 180 indicates upside down, and 270 indicates the image faces left.
content             The summary of recognized text blocks in the image.
height              The height of the image after algorithm correction.
width               The width of the image after algorithm correction.
orgHeight           The height of the original image.
orgWidth            The width of the original image.
prism_wnum          The number of recognized text blocks, which is the size of the prism_wordsInfo array.

Description of fields in the prism-wordsInfo text block array:

angle                The angle of the text block. This angle only affects width and height. When the angle is -90, 90, -270, or 270, you need to swap the width and height values.
height               The height of the text block.
width                The width of the text block.
pos                  The coordinates of the four corners of the bounding rectangle of the text block, arranged clockwise: top-left, top-right, bottom-right, bottom-left. When NeedRotate is set to true and the outer-level angle is not 0, the coordinates are accurate only after the image is corrected based on the angle.
word                 The text content of the text block.
tableId              This field exists when OutputTable is set to true and the text block is inside a table. tableId indicates the ID of the table.
tableCellId          This field exists when OutputTable is set to true and the text block is inside a table. It indicates the ID of the cell in the table.

charInfo single character information:

word                  The single character text.
x                     The x-coordinate of the top-left corner of the character.
y                     The y-coordinate of the top-left corner of the character.
w                     The width of the character.
h                     The height of the character.

Description of fields in the prism-tablesInfo table array:

tableId            The table ID, which corresponds to the tableId in prism_wordsInfo.
xCellSize          The number of cells along the horizontal axis in the table.
yCellSize          The number of cells along the vertical axis in the table.

cellInfos cell information, which contains the spatial topology of the cell within the entire table:

tableCellId        The cell ID in the table, which corresponds to the tableCellId in prism_wordsInfo.
word               The text in the cell.
xsc                Abbreviation for xStartCell. Indicates the starting cell position along the horizontal axis. The first cell has a value of 0.
xec                Abbreviation for xEndCell. Indicates the ending cell position along the horizontal axis. The first cell has a value of 0. If both xsc and xec are 0, the text occupies one cell along the horizontal axis and is in the first cell.
ysc                Abbreviation for yStartCell. Indicates the starting cell position along the vertical axis. The first cell has a value of 0.
yec                Abbreviation for yEndCell. Indicates the ending cell position along the vertical axis. The first cell has a value of 0.
pos                The cell position. The coordinates are arranged clockwise by the four corners of the cell: top-left XY, top-right XY, bottom-right XY, and bottom-left XY.