Model description
Model Name: cyrillic-htr-model
Model Type: Transformer-based OCR (TrOCR)
Base Model: microsoft/trocr-large-handwritten
Purpose: Handwritten text recognition
Languages: Cyrillic
License: Apache 2.0
This model is a fine-tuned version of the microsoft/trocr-large-handwritten model, specialized for recognizing handwritten cyrillic text. At the moment it has been trained on the dataset (number of pages 740) from 17th to 20th centuries.
Model Architecture
The model is based on a Transformer architecture (TrOCR) with an encoder-decoder setup:
- The encoder processes images of handwritten text.
- The decoder generates corresponding text output.
Intended Use
This model is designed for handwritten text recognition and is intended for use in:
- Document digitization (e.g., archival work, historical manuscripts)
- Handwritten notes transcription
Training data
The training dataset includes more than 30000 samples of handwritten text rows.
Evaluation
The model was evaluated on test dataset. Below are key metrics:
Character Error Rate (CER): 8
Test Dataset Description: size ~33 400 text rows
How to Use the Model
You can use the model directly with Hugging Face’s pipeline function or by manually loading the processor and model.
from transformers import TrOCRProcessor, VisionEncoderDecoderModel
from PIL import Image
# Load the model and processor
processor = TrOCRProcessor.from_pretrained("Kansallisarkisto/cyrillic-htr-model/processor")
model = VisionEncoderDecoderModel.from_pretrained("Kansallisarkisto/cyrillic-htr-model")
# Open an image of handwritten text
image = Image.open("path_to_image.png")
# Preprocess and predict
pixel_values = processor(image, return_tensors="pt").pixel_values
generated_ids = model.generate(pixel_values)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(generated_text)
Limitations and Biases
The model was trained primarily on handwritten text that uses basic Cyrillic characters.
Future Work
Potential improvements for this model include:
- Expanding training data: Incorporating more diverse handwriting styles and languages.
- Optimizing for specific domains: Fine-tuning the model on domain-specific handwriting.
Citation
If you use this model in your work, please cite it as:
@misc{cyrillic_htr_model_2025,
author = {Kansallisarkisto},
title = {Cyrillic HTR Model: Handwritten Text Recognition},
year = {2025},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/Kansallisarkisto/cyrillic-htr-model/}},
}
Model Card Authors
Author: Kansallisarkisto