Skip to content

DocRes Engine

Guide to using the DocRes image restoration engine.

Overview

The DocResEngine provides direct access to document image restoration capabilities using the DocRes model. Use it for standalone image enhancement or as part of the parsing pipeline.

Key Features

  • 6 Restoration Tasks: Comprehensive document enhancement
  • GPU Acceleration: CUDA support for faster processing
  • Flexible Input: Images or PDFs
  • Detailed Metadata: Processing information returned

Basic Usage

from doctra import DocResEngine

# Initialize engine
engine = DocResEngine(device="cuda")

# Restore image
restored_img, metadata = engine.restore_image(
    image="document.jpg",
    task="appearance"
)

# Save result
restored_img.save("restored.jpg")

Restoration Tasks

Task Description Use Case
appearance General enhancement Most documents
dewarping Fix perspective Scanned at angle
deshadowing Remove shadows Poor lighting
deblurring Reduce blur Motion/focus issues
binarization B&W conversion Clean text
end2end Full pipeline Severe degradation

PDF Restoration

engine = DocResEngine(device="cuda")

restored_pdf = engine.restore_pdf(
    pdf_path="low_quality.pdf",
    output_path="enhanced.pdf",
    task="appearance",
    dpi=300
)

See Also