Skip to content
  • Pricing
pagantibet · PyLaia · Published April 24, 2025

PaganTibet Ume 3

Text Recognition

Description

PaganTibet Ume 3 was developed as part of the PaganTibet project (pagantibet.com) at EPHE-PSL Paris, with funding from the European Union (ERC, Reconstructing the Pagan Religion of Tibet (2023-2028), 101097364). The model was trained on a dataset of 725 images from a diverse collection of Tibetan-language manuscripts, with 653 images used for training and 72 for validation. Most images consist of photographs of two manuscript pages. This model is an earlier iteration of PaganTibet Ume 5 (Transkribus Model ID: 443525).

Try this model

Drag an image here

Select a file...

PNG or JPG up to 10 Mb

Wolpi
AI Assistant

By uploading an image, you accept our terms and privacy policy.

Use this modelOpen in Transkribus
Low error rate6.26% CER

Character Error Rate (CER) measures the percentage of characters incorrectly recognised. Lower is better. This model scored 6.26% on its validation set. As a rule of thumb, a CER below 10% is considered good for most handwritten material. This is a larger model trained on diverse material, which generally makes it more robust across different handwriting styles. That said, larger training sets also make it harder to push the CER down further.

Measured on the model's own validation data. Results on your documents may differ depending on handwriting style, document condition, language, and how closely your material resembles the training data.

Words21,470
Lines5,064
Training Pages653
Model ID328353
Languages
Tibetan
Centuries
8th c.9th c.10th c.11th c.12th c.13th c.14th c.15th c.16th c.17th c.18th c.19th c.20th c.21st c.