R Client for the 'PaddleOCR' Cloud API

An R client for the 'PaddleOCR' cloud service API < https://www.paddleocr.ai/latest/en/version3.x/inference_deployment/serving/paddleocr_official_api/overview.html>. Submit images, PDFs, or URLs for OCR processing using models like 'PaddleOCR-VL-1.6'. Supports job submission, polling, result retrieval, automatic image download, and streaming PDF-to-markdown conversion with batch processing.

PaddleOCR

An R client for the PaddleOCR cloud API. Submit images, PDFs, or URLs for OCR processing and get structured Markdown output with automatic image downloading.

Installation

You can install the development version of PaddleOCR from GitHub:

# install.packages("devtools")
devtools::install_github("xiaoluolorn/PaddleOCR")

Setup

Set your PaddleOCR API token as an environment variable:

Sys.setenv(PADDLE_OCR_TOKEN = "your_api_token_here")

Or add it to your .Renviron file for persistence:

PADDLE_OCR_TOKEN=your_api_token_here

Quick Start

OCR a single image or URL

library(PaddleOCR)

# From a local file
result <- paddle_ocr("document.png")

# From a URL
result <- paddle_ocr("https://example.com/document.jpg")

# View results
result$markdown_files
result$page_count

OCR a PDF document

# Convert PDF to Markdown (requires pdftools)
result <- pdf_to_markdown_with_paddle("paper.pdf")

# The combined Markdown is saved and returned
cat(result$combined_markdown_text)

Interrupted runs resume automatically by reusing rendered page images, completed Markdown files, and submitted OCR job IDs. To run up to eight OCR jobs concurrently on the PaddleOCR service:

result <- pdf_to_markdown_with_paddle("paper.pdf", workers = 8)

Batch process PDFs

results <- batch_pdf_to_markdown_with_paddle(
  pdf_dir = "papers/",
  output_root = "papers/paddle_output"
)

Low-level API access

# Submit a job
job_id <- submit_paddle_job("image.png", token = "your_token")

# Poll for completion
jsonl_url <- poll_paddle_job(job_id, token = "your_token")

# Process the result
result <- process_paddle_jsonl_result(jsonl_url, output_dir = "output")

Features

• Simple API: paddle_ocr() handles everything from submission to result saving
• File & URL support: Submit local files or remote URLs
• PDF processing: Streaming page-by-page PDF rendering and OCR
• Batch processing: Process entire directories of PDFs
• Auto image download: Inline Markdown images and layout images are saved automatically
• Configurable: Model selection, orientation classification, unwarping, chart recognition

API Options

Dependencies

• Required: httr, jsonlite
• Optional: pdftools (for PDF processing)

License

MIT © PaddleOCR authors