README.md

metadata

title: PDF Layout Extractor
emoji: 📄
colorFrom: blue
colorTo: purple
sdk: docker
app_file: app.py
pinned: false

PDF Layout Extraction Tool

A powerful tool for extracting figures, tables, annotated layouts, and markdown text from scientific PDFs using DocLayout-YOLO.

Features

• REST API endpoints for programmatic access
• Real-time progress tracking with visual progress bar
• Multiple processing modes: Images only, Markdown only, or Both
• Background processing - upload files and track progress via API
• Modern web UI with dark/light theme
• GPU/CPU support - automatically detects available hardware

API Endpoints

Base URL: https://saifisvibin-volaris-pdf-tool.hf.space

• POST /api/predict - Recommended: Synchronous PDF extraction (returns complete results immediately)
• POST /api/upload - Upload PDFs for async processing (returns task_id)
• GET /api/progress/<task_id> - Get processing progress (0-100%)
• GET /api/pdf-list - List all processed PDFs
• GET /api/pdf-details/<pdf_stem> - Get details for a processed PDF
• GET /api/device-info - Get GPU/CPU device information
• GET /api/docs - Interactive API documentation
• GET /output/<path> - Download processed files (PDFs, images, markdown)

Example API Usage

Simple Synchronous Extraction (Recommended)

import requests

# Upload and get results immediately
files = {'file': open('document.pdf', 'rb')}
response = requests.post('https://saifisvibin-volaris-pdf-tool.hf.space/api/predict', files=files)
result = response.json()

print(f"Text: {result['text']}")
print(f"Figures: {len(result['figures'])}")
print(f"Tables: {len(result['tables'])}")

Async Processing with Progress

import requests
import time

# Upload a PDF (async)
files = {'files[]': open('document.pdf', 'rb')}
data = {'extraction_mode': 'both'}  # or 'images' or 'markdown'
response = requests.post('https://saifisvibin-volaris-pdf-tool.hf.space/api/upload', files=files, data=data)
task_id = response.json()['task_id']

# Poll for progress
while True:
    progress = requests.get(f'https://saifisvibin-volaris-pdf-tool.hf.space/api/progress/{task_id}').json()
    print(f"Progress: {progress['progress']}% - {progress['message']}")
    if progress['status'] == 'completed':
        break
    time.sleep(0.5)

# Get results
results = progress['results']

Processing Modes

1. Images Only - Extracts figures and tables with layout detection
2. Markdown Only - Extracts text content as markdown
3. Both - Extracts both images and markdown content

Output Structure

Each processed PDF creates a directory with:

• *_content_list.json - Metadata for extracted figures/tables
• *_layout.pdf - Annotated PDF with layout boxes
• *.md - Markdown export (if enabled)
• figures/ - Extracted figure images
• tables/ - Extracted table images

Built With

• DocLayout-YOLO - Layout detection
• PyMuPDF - PDF processing
• Flask - Web framework