Tigran83/russian-university-math-handwriting
A dataset of Russian handwritten mathematics documents containing handwritten solutions together with corresponding typed, LaTeX and Markdown versions, used for OCR, image-to-LaTeX conversion and mathematical reasoning tasks.
Timeline (last 90 days)
A dataset named Tigran83/russian-university-math-handwriting exists on Hugging Face, dedicated to recognition and reasoning tasks on Russian handwritten mathematics documents.
The dataset's free sample contains 50 math problems, covering 19 branches of mathematics and 49 different topics, with 117 original handwritten pages in total, of which 29 problems are accompanied by figures, and about 1,600 mathematical formulas in total.
The full commercial dataset contains 1,740 math problems, 3,450 pages of handwritten solutions, 50,603 mathematical formulas, and 927 mathematical figures, covering 19 branches and 127 topics, with a data volume of about 2.2 GB.
The full commercial dataset is priced at US$45,000, and purchase by mathematical branch is also supported.
The commercial license is a non-exclusive AI training and evaluation license that permits commercial model deployment and release of trained model weights, but does not permit redistribution of the original data.
The 50-problem evaluation sample in the free repository is governed by the terms of SAMPLE_LICENSE.md, and unlimited redistribution or use for commercial training is not permitted.
The handwritten and typed versions in this dataset are aligned by problem, not required to correspond line by line, so it is suitable for training whole-page-to-text generation models, but line-level detection or reading-order recovery tasks require additionally generated line-level annotations.
The dataset page explicitly lists four target application scenarios: handwritten math OCR, image-to-LaTeX conversion, LLM mathematical reasoning, and multimodal document understanding.
This dataset is Russian-only, the resolution and writing style of the handwritten pages may be inconsistent, and the samples are for evaluation and research only and may not be redistributed.
All Facts (9)
A dataset named Tigran83/russian-university-math-handwriting exists on Hugging Face, dedicated to recognition and reasoning tasks on Russian handwritten mathematics documents.
50%UnverifiedThe dataset's free sample contains 50 math problems, covering 19 branches of mathematics and 49 different topics, with 117 original handwritten pages in total, of which 29 problems are accompanied by figures, and about 1,600 mathematical formulas in total.
50%UnverifiedThe full commercial dataset contains 1,740 math problems, 3,450 pages of handwritten solutions, 50,603 mathematical formulas, and 927 mathematical figures, covering 19 branches and 127 topics, with a data volume of about 2.2 GB.
50%UnverifiedThe full commercial dataset is priced at US$45,000, and purchase by mathematical branch is also supported.
50%UnverifiedThe commercial license is a non-exclusive AI training and evaluation license that permits commercial model deployment and release of trained model weights, but does not permit redistribution of the original data.
50%UnverifiedThe 50-problem evaluation sample in the free repository is governed by the terms of SAMPLE_LICENSE.md, and unlimited redistribution or use for commercial training is not permitted.
50%UnverifiedThe handwritten and typed versions in this dataset are aligned by problem, not required to correspond line by line, so it is suitable for training whole-page-to-text generation models, but line-level detection or reading-order recovery tasks require additionally generated line-level annotations.
50%UnverifiedThe dataset page explicitly lists four target application scenarios: handwritten math OCR, image-to-LaTeX conversion, LLM mathematical reasoning, and multimodal document understanding.
50%UnverifiedThis dataset is Russian-only, the resolution and writing style of the handwritten pages may be inconsistent, and the samples are for evaluation and research only and may not be redistributed.
50%