ocr_functions.py

from azure.core.credentials import AzureKeyCredential
from azure.ai.formrecognizer import DocumentAnalysisClient
from io import BytesIO
from helpers import format_polygon, Logger
import logging
import os
import sys 

logging.basicConfig(filename='app.log', level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
sys.stdout = Logger("output.log")

endpoint = os.environ['AZURE_API_ENDPOINT']
key  = os.environ['AZURE_API_KEY']

def detect_document(content):
    document_analysis_client = DocumentAnalysisClient(
        endpoint=endpoint, credential=AzureKeyCredential(key)
    )

    with open(content.name, "rb") as f:
        poller = document_analysis_client.begin_analyze_document(
            "prebuilt-document", document=f
        )
    result = poller.result()

    text_content = result.content
    logging.info(result.content)
    print(text_content)

    pair_content = "----Key-value pairs found in document----\n"
    for kv_pair in result.key_value_pairs:
        if kv_pair.key and kv_pair.value:
            pair_content += "Key '{}' with Value '{}' \n".format(
                kv_pair.key.content,
                kv_pair.value.content
            )
    logging.info(pair_content)
    print(pair_content)
    document_content = "----Lines found in document----\n"
    for page in result.pages:
        for line_idx, line in enumerate(page.lines):
            document_content += "...Line # {} text '{}' within bounding polygon '{}' \n".format(
                    line_idx,
                    line.content,
                    format_polygon(line.polygon),
                )
    logging.info(document_content)
    print(document_content)
    table_content = "----Tables found in document----\n"
    for table_idx, table in enumerate(result.tables):
        table_content += "Table # {} has {} rows and {} columns\n".format(
            table_idx, table.row_count, table.column_count
        )
        for cell in table.cells:
            table_content += "...Cell[{}][{}] has content '{}'\n".format(
                cell.row_index,
                cell.column_index,
                cell.content,
            )
    logging.info(table_content)
    print(table_content)
    name = content.name.split('\\')[-1]
    name = name.split("/")[-1]
    name = name.split('.')[0]
    return (text_content, pair_content, document_content, table_content, name)

def detect_image(content):
    document_analysis_client = DocumentAnalysisClient(
        endpoint=endpoint, credential=AzureKeyCredential(key)
    )
    byte_stream = BytesIO()
    content.save(byte_stream, format='PNG') # or 'JPEG', 'BMP', etc. depending on your image
    byte_stream.seek(0)  # reset pointer back to the start of the stream
    poller = document_analysis_client.begin_analyze_document(
        "prebuilt-read", document=byte_stream
    )

    result = poller.result()
    logging.info(result.content)
    print(result.content)
    return(result.content)