Textract-Overlayer

amazon-textract-overlayer 提供帮助在文档上叠加边界框的功能。

安装

> python -m pip install amazon-textract-overlayer

请确保您的环境已通过配置文件、环境变量或附加的角色设置 AWS 凭据。 (https://docs.aws.amazon.com/cli/latest/userguide/cli-chap-configure.html)

示例

提供的主要方法是 get_bounding_boxes，它根据传入的 Textract_Type 返回边界框。主要来自 amazon-textract 命令，该命令来自 amazon-textract-helper 包。

这将返回 WORD 和 CELL 数据类型的边界框。

from textractoverlayer.t_overlay import DocumentDimensions, get_bounding_boxes
from textractcaller.t_call import Textract_Features, Textract_Types, call_textract

doc = call_textract(input_document=input_document, features=features)
# image is a PIL.Image.Image in this case
document_dimension:DocumentDimensions = DocumentDimensions(doc_width=image.size[0], doc_height=image.size[1])
overlay=[Textract_Types.WORD, Textract_Types.CELL]

bounding_box_list = get_bounding_boxes(textract_json=doc, document_dimensions=document_dimension, overlay_features=overlay)

图像边界框的实际叠加绘制位于 amazon-textract 命令中，该命令来自 amazon-textract-helper 包，外观如下

from PIL import Image, ImageDraw

image = Image.open(input_document)
rgb_im = image.convert('RGB')
draw = ImageDraw.Draw(rgb_im)

# check the impl in amazon-textract-helper for ways to associate different colors to types
for bbox in bounding_box_list:
    draw.rectangle(xy=[bbox.xmin, bbox.ymin, bbox.xmax, bbox.ymax], outline=(128, 128, 0), width=2)

rgb_im.show()

要在 PDF 文档中绘制边界框，可以使用以下代码

import fitz

# for local stored files
file_path = "<<replace with the local path to your pdf file>>"
doc = fitz.open(file_path)
# for files stored in S3 the streaming object can be used
# doc = fitz.open(stream="<<replace with stream_object_variable>>", filetype="pdf")

# draw boxes
for p, page in enumerate(doc):
    p += 1
    for bbox in bounding_box_list:
        if bbox.page_number == p:
            page.draw_rect(
                [bbox.xmin, bbox.ymin, bbox.xmax, bbox.ymax], color=(0, 1, 0), width=2
            )

# save file locally 
doc.save("<<local path for output file>>")