FORMAT_CONVERSION_MAXCOMPUTE/format_convert/convert_tree.py

import io
import cv2
from PIL import Image
import numpy as np
from format_convert.convert_image import image_preprocess
from format_convert.utils import add_div, judge_error_code, get_table_html, sort_object


class _Document:
    def __init__(self, doc_path):
        self.doc_path = doc_path
        # Document's child -> Page
        self.children = []
        self.error_code = None

    def add_child(self, child):
        if child.error_code is None:
            self.children.append(child)
        else:
            self.error_code = child.error_code

    def get_html(self):
        if self.error_code is not None:
            return self.error_code

        html_text = ""
        for child in self.children:
            # 先调用get_html才能更新error_code
            child_html_text = child.get_html()
            print("Document", self.error_code, child.error_code, type(child), child.page_no)
            if child.error_code is not None:
                self.error_code = child.error_code
                return self.error_code
            else:
                html_text += child_html_text
        return [html_text]


class _Page:
    def __init__(self, page, page_no):
        self.page = page
        self.page_no = page_no
        # Page's child -> Image, Table, Sentence
        self.children = []
        self.error_code = None
        # objs in tables
        self.in_table_objs = set()

    def add_child(self, child):
        if child.error_code is None:
            self.children.append(child)
        else:
            self.error_code = child.error_code

    def get_html(self):
        if self.error_code is not None:
            return ""

        html_text = ""
        self.children = sort_object(self.children)
        for child in self.children:
            # 先调用get_html才能更新error_code
            child_html_text = child.get_html()
            print("Page", self.error_code, child.error_code, type(child))
            if child.error_code is not None:
                self.error_code = child.error_code
                return ""
            else:
                html_text += child_html_text
        return html_text


class _Image:
    def __init__(self, content, path):
        self.content = content
        self.path = path
        # 来源
        self.is_from_pdf = False
        # 位置
        self.x = 0
        self.y = 0
        # 识别结果
        self.otr_result = None
        self.ocr_result = None
        # Image's child -> Table, Sentence
        self.children = []
        self.error_code = None
        # objs in tables
        self.in_table_objs = set()

    def add_child(self, child):
        if child.error_code is None:
            self.children.append(child)
        else:
            self.error_code = child.error_code

    def get_html(self):
        # 将Image转为Sentence,table
        self.convert()
        print("Image", self.error_code)
        if self.error_code is not None:
            return ""

        html_text = ""
        self.children = sort_object(self.children)
        for child in self.children:
            # 先调用get_html才能更新error_code
            child_html_text = child.get_html()
            print("Image", self.error_code, child.error_code, type(child))
            if child.error_code is not None:
                self.error_code = child.error_code
                return ""
            else:
                html_text += child_html_text
        return html_text

    def get_text(self):
        return

    def convert(self):
        # 二进制转numpy
        image_np = Image.open(io.BytesIO(self.content))
        image_np = cv2.cvtColor(np.asarray(image_np), cv2.COLOR_RGB2BGR)
        text, column_list, outline_points, is_table = image_preprocess(image_np,
                                                                       self.path,
                                                                       use_ocr=True)
        print("is_table", is_table)
        for t in text:
            print(t)
        if judge_error_code(text):
            self.error_code = text
            return
        if is_table:
            tables, in_objs = text
            self.in_table_objs = in_objs
            for table in tables:
                self.add_child(_Table(table["table"], table["bbox"]))
        else:
            self.add_child(_Sentence(text))


class _Table:
    def __init__(self, content, bbox):
        self.content = content
        self.bbox = bbox
        self.x = bbox[0]
        self.y = bbox[1]
        self.shape = (len(content), len(content[0]))
        self.error_code = None

    def get_html(self):
        if self.error_code is not None:
            return ""

        # 将二维数组转为html table
        html_text = get_table_html(self.content)
        return html_text


class _Sentence:
    def __init__(self, content):
        self.content = content
        # 位置
        self.x = 0
        self.y = 0
        self.error_code = None

    def get_html(self):
        if self.error_code is not None:
            return ""
        return add_div(self.content)


class TextBox:
    def __init__(self, bbox, text):
        self.bbox = bbox
        self.text = text

    def get_text(self):
        return self.text


class TableLine:
    def __init__(self, bbox):
        self.bbox = bbox