RustOCR

A fast Rust CLI wrapper for EasyOCR providing OCR capabilities with 80+ language support.

Features

• 🚀 Fast CLI interface built with Rust
• 🌍 80+ languages supported (via EasyOCR)
• 🎯 GPU acceleration support (CUDA/MPS)
• 📊 Multiple output formats: JSON, text, detailed
• 🔧 Easy to use with sensible defaults

Prerequisites

Before using RustOCR, ensure you have:

1. Python 3.7+ installed

2. EasyOCR Python library:

   pip install easyocr
   

   Note: If you encounter import errors with EasyOCR (e.g., bidi module issues), try reinstalling:

   pip install --upgrade --force-reinstall python-bidi easyocr
   

3. Rust toolchain (for building from source):

   curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
   

Installation

Build from source

cd /home/user/EasyOCR/rustOCR
cargo build --release

The binary will be available at target/release/rustocr.

Install system-wide (optional)

cargo install --path .

Usage

Basic Examples

# Simple OCR on an English image
./target/release/rustocr -i image.jpg

# OCR with Chinese and English
./target/release/rustocr -i chinese.jpg -l ch_sim,en

# Use CPU instead of GPU
./target/release/rustocr -i document.png -g false

# Get simple text output (no bounding boxes)
./target/release/rustocr -i image.jpg -d 0 -o text

# Detailed human-readable output
./target/release/rustocr -i image.jpg -o detailed

Command-Line Options

Options:
  -i, --input <FILE>         Input image file path
  -l, --languages <LANGS>    Languages to recognize (comma-separated) [default: en]
  -g, --gpu <BOOL>          Enable GPU acceleration [default: true]
  -d, --detail <LEVEL>       Detail level: 0 (text only) or 1 (full) [default: 1]
  -o, --output <FORMAT>      Output format: json, text, or detailed [default: json]
  -h, --help                Print help
  -V, --version             Print version

Supported Languages

EasyOCR supports 80+ languages including:

• Latin: English (en), French (fr), German (de), Spanish (es), Portuguese (pt), Italian (it), etc.
• Chinese: Simplified (ch_sim), Traditional (ch_tra)
• Japanese (ja), Korean (ko), Thai (th)
• Arabic (ar), Persian (fa), Urdu (ur)
• Cyrillic: Russian (ru), Ukrainian (uk), Bulgarian (bg), etc.
• Devanagari: Hindi (hi), Marathi (mr), Nepali (ne)
• And many more...

See thefull list of supported languages.

Output Formats

JSON (default)

[
  {
    "bbox": [[10, 20], [100, 20], [100, 50], [10, 50]],
    "text": "Hello World",
    "confidence": 0.9872
  }
]

Text

Hello World
Sample Text

Detailed

--- Result 1 ---
Text: Hello World
Confidence: 0.9872
Bounding Box: [[10, 20], [100, 20], [100, 50], [10, 50]]

Performance Tips

• GPU: For best performance, use GPU mode (default). Requires CUDA or MPS support.
• CPU: Use -g false if you don't have GPU support. Performance will be slower but still functional.
• Batch Processing: For multiple images, call the binary in a loop or script.

Troubleshooting

"Failed to import easyocr"

Make sure EasyOCR is installed:

pip install easyocr

GPU-related errors

If you encounter GPU errors, try CPU mode:

./target/release/rustocr -i image.jpg -g false

"Input file does not exist"

Ensure the image path is correct and the file exists.

Architecture

RustOCR uses PyO3 to bridge Rust and Python, calling the EasyOCR library at runtime:

┌─────────────┐
│  Rust CLI   │ (Argument parsing, error handling)
│   (clap)    │
└──────┬──────┘
       │
       ↓
┌─────────────┐
│    PyO3     │ (Rust ↔ Python bridge)
└──────┬──────┘
       │
       ↓
┌─────────────┐
│  EasyOCR    │ (Python library)
│  (PyTorch)  │
└─────────────┘

License

Apache-2.0 (matching the parent EasyOCR project)

Credits

• EasyOCR2 - The underlying OCR engine
• Built with PyO3 and Clap