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