Transformer-Based OCR for Russian Cyrillic Word Recognition

This repository by Chebart provides a dataset and implementation for Russian word optical character recognition (OCR) using Transformer architectures. Updated in November 2023, the project utilizes PyTorch and HuggingFace Transformers to process Cyrillic text. The dataset focuses specifically on the recognition of individual Russian words rather than full documents.

Use Cases

• Training Transformer-based models for Russian Cyrillic text recognition
• Fine-tuning HuggingFace Vision Encoder-Decoder models for OCR tasks
• Benchmarking Russian word recognition accuracy in PyTorch environments

Strengths

• Focuses specifically on Russian Cyrillic character sets
• Compatible with HuggingFace Transformers library
• Built on the PyTorch framework for deep learning

Limitations

• Total record count and dataset size are unknown
• Lack of documentation regarding the primary data source or annotation method
• No specified license for commercial or research use

Provenance

Source

GitHub repository by user Chebart

Freshness

Last updated November 25, 2023.

Geography

["Russia"]