Installation.md

Installing OCR dependencies

• Install the dependencies as required for your inference
• Recommended to use conda / virtualenv for each type of inference to avoid version lock conflicts.

---

End-to-end solutions

Tessarect OCR

• If you're on Linux, best option: sudo apt-get install tesseract-ocr-all
• If you're on Windows, use the latest installer from here and install all required languages by choosing that option during installation.
• Ensure it's properly installed by typing tesseract on command line.
• Then: pip install pytesseract
• Optional: To manually get language files, go here.

Clova.AI EasyOCR

• Ensure PyTorch is installed (Optional: with CUDA+CuDNN for GPU)
• Install the library:
  • For stable: pip install easyocr
  • For latest: pip install git+https://github.com/JaidedAI/EasyOCR
• Sample config: easy_ocr.json
• Languages supported | Parameters allowed

Google Vision OCR

• pip install google-cloud-vision
• Sample config: google_ocr.json
• Ensure to set service_account_json path in config if not using GCP
• Pricing

---

Detection

CRAFT

• pip install craft-text-detector
• To run on GPU, install CUDA & CuDNN with corresponding PyTorch.
• Sample config: craft+tesseract.json

OpenCV EAST Detector

• Uses OpenCV's DNN Module (CPU Only)
• Sample config: east+tesseract.json

DB Text Detector

• Ensure the PyTorchOCR repo is cloned at libs/PyTorchOCR.
• Sample config: db+tesseract.json

---

Recognition

Custom ClovaAI

• Ensure the Clova AI repo is cloned at libs/clova_ai_recognition.
• (...TODO: Add more )
• Sample config: craft+clova_ai.json