Datasets:
Size:
1M<n<10M
ArXiv:
Tags:
Document_Understanding
Document_Packet_Splitting
Document_Comprehension
Document_Classification
Document_Recognition
Document_Segmentation
DOI:
License:
File size: 2,134 Bytes
165da3c | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 | # Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
# SPDX-License-Identifier: CC-BY-NC-4.0
import os
from pathlib import Path
from typing import List
from PIL import Image
import fitz
from loguru import logger
class AssetWriter:
"""Writes document assets (PDF, images, OCR text) to disk."""
def __init__(self, output_base_path: str = ".data/rvl-cdip-nmp-hf/rvl-cdip-nmp-assets"):
self.output_base_path = Path(output_base_path)
def save_document_assets(
self,
doc_type: str,
doc_name: str,
filename: str,
pdf_bytes: bytes,
text_pages: List[str]
):
"""Save all assets for a document: original PDF, page images, and OCR text."""
# Create directory structure: {doc_type}/{filename}/
doc_dir = self.output_base_path / doc_type / filename
original_dir = doc_dir / "original"
pages_dir = doc_dir / "pages"
original_dir.mkdir(parents=True, exist_ok=True)
# Save original PDF
original_pdf_path = original_dir / filename
with open(original_pdf_path, 'wb') as f:
f.write(pdf_bytes)
# Extract and save page images
pdf_doc = fitz.open(stream=pdf_bytes, filetype="pdf")
for page_num in range(len(pdf_doc)):
page = pdf_doc[page_num]
page_num_str = f"{page_num + 1:04d}"
# Create page directory
page_dir = pages_dir / page_num_str
page_dir.mkdir(parents=True, exist_ok=True)
# Save page image
pix = page.get_pixmap(dpi=300)
img_path = page_dir / f"page-{page_num_str}.png"
pix.save(str(img_path))
# Save OCR text
if page_num < len(text_pages):
text_path = page_dir / f"page-{page_num_str}-textract.md"
with open(text_path, 'w', encoding='utf-8') as f:
f.write(text_pages[page_num])
pdf_doc.close()
logger.debug(f"Saved assets for {doc_type}/{doc_name}")
|