ScrabbleGAN-YeniseiGovReports-HWR: 788,436 Synthetic Historical Russian Handwriting Images

Yenisei Governorate reports from historical archives form the basis for this collection of 788,436 synthetically generated images of pre-reform Russian words with corresponding text transcriptions. The dataset was created by author sherstpasha using a ScrabbleGAN generative model adapted to mimic the historical handwriting style of these reports. It was last updated on April 10, 2026.

Use Cases

• Training optical character recognition (OCR) models for historical Russian script based on the synthetic handwriting images.
• Benchmarking generative adversarial networks (GANs) for document synthesis based on the described ScrabbleGAN methodology.
• Developing data augmentation pipelines for historical document digitization projects based on the synthetic generation approach.
• Studying visual features of archival scans, such as background and noise patterns, mentioned in the description.

Strengths

• Contains 788,436 synthetically generated image-text pairs, providing a substantial volume of training data.
• Generation process considered specific visual characteristics like background and noise from archival scans of Yenisei Governorate reports.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the textual data is unknown, which may limit suitability assessment.
• The dataset is synthetic, which may introduce artifacts not present in genuine historical handwriting.

Provenance

Source

huggingface, author sherstpasha

Collection Method

Synthetically generated using a ScrabbleGAN model adapted to historical handwriting style.

Freshness

Last updated 2026-04-10 17:48:51; freshness should be verified.

Geography

Likely associated with the historical Yenisei Governorate (Russia).