Namgyal Manuscript Collection Datasets

These are the official datasets created for the Tibetan Manuscript Project Vienna (TMPV) in the years 2023 and 2024. These datasets contain:

• OCR datasets (line image - line label pairs) created from the PageXML annotations
• PageXML (Transkribus) annotations in Unicode and Wylie
• PageXML Layout annotations (lines, images, captions, margins) used for image segmentation training
• OCR models (PyTorch checkpoints and ONNX model files)