Dataset Viewer
Auto-converted to Parquet Duplicate
image
imagewidth (px)
1.7k
1.7k
pdf_marker
int64
121
601
markdown
stringlengths
5.57k
6.67k
inference_info
stringclasses
1 value
121
ch’ang-shih 263 ch’ang-shih 敬史 Clerk. (1) LIAO: minor officials in both Northern and Southern Bureaus of Military Affairs (shu-mi yuán). (2) CH’ING: variant reference to Clerks called pi-jié-shih. P5, 12. 264 ch’ang-shih chi 常侍 HAN: Mounted Attendant-in-ordinary, an honorary office (chia-kuan) awarded to favored offici...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
140
135 612-626 (ch’in-wang). Then in 1635 submissive Mongols and Chinese were organized into 8 similarly designated Mongol Banners (meng-ku ch’i) and 8 similarly designated Chinese Banners (han-ch’in ch’i). The number of Mongol Banners increased greatly later in the dynasty, as more Mongol groups submitted to Ch’ing au...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
183
chiung-t’ai Court of the Imperial Stud (t’ai-p’u ssu ching, t’ai-p’u ching). 1329 chiung-t’ai 間臺 CH’ING: unofficial reference to the Court of the Imperial Stud (t’ai-p’u ssu). 1330 chó 洞 Beginning in the era of N-S Division very soon after the fall of Han, a term meaning impure used as an at least quasi-official des...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
274
269 2991-3010 i-t'ing (li-pu), from 1389 to 1396 only; previously called General Bureau (tsung-pu), later called Bureau of Ceremonies (chi-ch'ing-li ssu). (2) MING-CH'ING: unofficial reference to the Ministry of Rites. 2991 i-shàn 翊善 SUNG: Moral Mentor, one, rank 7b, in each Prince's Establishment (wang-fu) and ...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
288
283 3255-3274 kuan-chi she-jen subofficial functionaries (and possibly some eunuchs) in the Palace Administration (tien-chung sheng). P38. 3255 k'u-t'ing 庫應 CH'ING: unofficial reference to a Storehouse Commissioner-in-chief (ku ta-shi) in a Provincial Administration Commission (ch'eng-hsuan pu-cheng shih ssu). 3256 ...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
305
# kuo-tzu po-shih early Sung was specially charged with educating sons of the most eminent nobles and officials in the Confucian classics. Always closely affiliated with the Court of Imperial Sacrifices (t'ai-ch'ang ssu) and the Ministry of Rites (li-pu), and in Yuan subordinated to the Academy of Scholarly Worthies (...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
327
lou-k'o k'o 3833-3848 Hsiung-nu tribe in modern Shansi whose name, Lou-fan in Chinese transliteration, meant expert archer, or from the name of an early Han man who was renowned as a skilled archer. 3833 lou-k'o k'o 滿剌科 Water Clock Section, responsible for maintaining and teaching use of the palace water clocks and f...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
423
shen-hsing i-kuan ment (nei-wu fu); directed palace eunuchs and Shamansess (sa-man tai-tai) in offering prayers for the deceased Man-chu rulers. BH: office of shamanism. 5152 shen-hsing i-kuan 審判議官 SUNG: Consultant in the Review of Sentences, a duty assignment in the Ministry of Justice (hsiing-pu); status details no...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
427
shih-chang 5206-5223 422 signed to care for different chambers (shih) of the temples in which ancestral spirits of the imperial clan, the spirits of deceased Empresses, etc., were honored or worshipped; number and rank not specified. P28. 5206 shih-chang 市長 Variant of shih-ling (Market Director), sometimes used for le...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
528
523 6961-6979 to-slo-ling t’ing forces organized at the end of Han; see pa hsiao-wei (Eight Commandants). 6961 tsō-chūn 左軍 Left Army. (1) Throughout history a common designation for one of 3 or 5 military forces in battle array, others normally prefixed right and center (sometimes also front and rear). (2) HAN: ...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
595
yi-fu stored jade, pearls, and precious metal objects that had been presented to the King or collected in taxes, to be used as personal adornments or items of ritual use or decor by the King; the Storekeepers included 2 ranked as Senior Servicemen (shang-shi) and 4 as Ordinary Servicemen (chung-shi). CL: (chef de) mag...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]
601
yuan-ch'in prefixes. (2) T'ANG: Brokerage, a regional salt control office in a major market center, where the distribution of state-monopolized salt was supervised and salt taxes were collected; headed by a Brokerage Official (chih-yuan kuan); subordinate to a regional Directorate (chien) and ultimately to the Salt Mo...
[{"model_id": "deepseek-ai/DeepSeek-OCR-2", "model_name": "DeepSeek-OCR-2", "column_name": "markdown", "timestamp": "2026-08-13T22:40:48.730781", "prompt_mode": "free", "max_tokens": 8192}]

Document OCR using DeepSeek-OCR-2

This dataset contains markdown-formatted OCR results from images in bokane/hucker-ocr-bakeoff using DeepSeek-OCR-2.

Processing Details

Configuration

  • Image Column: image
  • Output Column: markdown
  • Dataset Split: train
  • Batch Size: 8
  • Max Model Length: 8,192 tokens
  • Max Output Tokens: 8,192
  • GPU Memory Utilization: 80.0%

Model Information

DeepSeek-OCR-2 is a 3B parameter vision-language model featuring Visual Causal Flow architecture for more human-like visual encoding. Building on DeepSeek-OCR v1, it offers enhanced document understanding with dynamic resolution up to (0-6)x768x768 + 1x1024x1024 patches.

Capabilities

  • LaTeX equations - Mathematical formulas preserved in LaTeX format
  • Tables - Extracted and formatted as HTML/markdown
  • Document structure - Headers, lists, and formatting maintained
  • Image grounding - Spatial layout and bounding box information
  • Complex layouts - Multi-column and hierarchical structures
  • Multilingual - Supports multiple languages

Dataset Structure

The dataset contains all original columns plus:

  • markdown: The extracted text in markdown format with preserved structure
  • inference_info: JSON list tracking all OCR models applied to this dataset

Usage

from datasets import load_dataset
import json

# Load the dataset
dataset = load_dataset("{{output_dataset_id}}", split="train")

# Access the markdown text
for example in dataset:
    print(example["markdown"])
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Column: {{info['column_name']}} - Model: {{info['model_id']}}")

Reproduction

This dataset was generated using the uv-scripts/ocr DeepSeek-OCR-2 vLLM script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/deepseek-ocr2-vllm.py \\
    bokane/hucker-ocr-bakeoff \\
    <output-dataset> \\
    --image-column image

Performance

  • Processing Speed: ~0.1 images/second
  • Processing Method: Batch processing with vLLM (2-3x speedup over sequential)

Generated with UV Scripts

Downloads last month
29