Tibetan OCR local app

This app is a free and open source offline App that can be installed on a local compter to run Tibetan OCR on batches of images (including in PDFs). It was developed by Eric Werner for the Buddhist Digital Resource Center.

Main features

The app can open one or multiple files and run OCR on them. It can export plain text or PageXML (a format it shares with Transkribus).

It can also optionally dewarp images as well as convert the output to Wylie.

Instead of providing one model that can handle all styles of Tibetan writing, we provide a few different models that we encourage users to experiment with to see what fits their data best.

The models it uses are based on transcriptions available online, from BDRC, ALL, Adarsha, and NorbuKetaka, as well as some transcriptions by MonlamAI and the author. The data was organized and processed in collaboration with MonlamAI, and parts of it can be made available on request.

Installation and running

Windows

1. Download and unzip https://github.com/buda-base/tibetan-ocr-app/releases/download/v0.1/bdrc_ocr_windows_x64_0.1.zip
2. Run main.exe
3. configure the models (see below)

OSX

This app does not have a package for MacOS on Intel processors at the moment (contributions welcome!). For Apple (M1, M2, etc.) processors:

1. Download and unzip

• for recent hardware https://github.com/buda-base/tibetan-ocr-app/releases/download/v0.1/bdrc_ocr_macos_arm64_0.1.zip
• for older x64 (Intel) processor https://github.com/buda-base/tibetan-ocr-app/releases/download/v0.1/bdrc_ocr_macos_x64_0.1.zip

2. Run the app
3. configure the models (see below)

From source (advanced users)

1. Make sure you have Git LFS installed
2. Clone the Github repository: git clone https://github.com/buda-base/tibetan-ocr-app.git
3. run git lfs pull to download all LFS files
4. Install dependencies with pip install -r requirements.txt (requires at least Python 3.10)
5. run pyside6-rcc resources.qrc -o resources.py
6. Run python main.py
7. configure the models (see below)

Configuring the models

1. Download and unzip https://github.com/buda-base/tibetan-ocr-app/releases/download/v0.1/bdrc_ocr_models_1.0.zip in a directory of your choice
2. Open the app, click on the setting icon, click on "import models" and select the ORCModels/ folder where you extracted the model zip file. Warning! Do not select one of its subfolders (like Woodblock/, etc.).
3. Then quit the app and run it again so that the models can be used.

Building distribution packages

1. pip install nuitka
2. (optional) install ccache to speed up the compilation (on OSX this can be done through homebrew)
3. run the nuitka command given in main.py that corresponds to your OS
4. zip the files in the corresponding build folder

Acknowledgements

Our gratitude goes to Jérémy Frère for the OSX packaging.