Skip to page content
SAVRN
Search Plan a project

Dataset · Image to text

synthetic-ocr-images-ko-1M

by Junyeong Song junyeong-nero/synthetic-ocr-images-ko-1M

This repository contains a synthetic OCR dataset generated by the Synthetic OCR Image Generator pipeline. It is intended for benchmarking and evaluating OCR or VLM systems on markdown-oriented document recognition tasks.

Rows—
Configurations—
Size114.8 GB
Licensecc-by-sa-3.0
AccessPublicly accessible
Monthly Downloads1

Dataset Card

By Junyeong Song, published under cc-by-sa-3.0, revision 5c72a08b6ed3.

Synthetic OCR Dataset

Work in progress. This dataset is generated and uploaded shard by shard. 515,000 of 1,000,000 samples are available so far (train 463,500 / test 51,500, as of 2026-09-30 21:55 UTC). This card is regenerated when the last shard is uploaded, so the counts below are partial.

This repository contains a synthetic OCR dataset generated by the Synthetic OCR Image Generator pipeline. It is intended for benchmarking and evaluating OCR or VLM systems on markdown-oriented document recognition tasks.

At a Glance

Included Splits

  • train: 463,500 samples
  • test: 51,500 samples

Each example is generated automatically and published with split-aware metadata for reproducible evaluation workflows.

Generation Configuration

  • Generated at (UTC): 2026-09-30T21:55:26.402900Z
  • Template: random
  • Template family: auto
  • Min template complexity: auto
  • Max template complexity: auto
  • Template config dir: auto
  • Markdown renderer: playwright
  • Style profile: balanced
  • Coverage targets: auto
  • Novelty window: 80
  • Novelty threshold: 0.95
  • Novelty max attempts: 4
  • Similar char ratio: auto
  • Similarity DB path: auto
  • Formula source mode: mixed
  • Formula dataset path: auto
  • Formula source weights (dataset/random/synthetic): 0.45/0.3/0.25
  • Add noise: auto
  • Add blur: auto
  • Seed: 42
  • Distribution profile: auto
  • Train ratio: 0.9
  • Test ratio: 0.1

Text Source and Attribution

Document text is derived from: Korean WikiText (https://github.com/lovit/kowikitext), CC BY-SA 3.0

Sample Distribution

Counts are computed from the published metadata.jsonl. Every column below is also available per sample for filtering.

document_family

Value Count Share
technical 150,100 29.1%
business 106,040 20.6%
forms 91,116 17.7%
academic 72,030 14.0%
operations 51,800 10.1%
sections 43,914 8.5%

block_type

Value Count Share
table 634,855 22.7%
paragraph 488,209 17.5%
checklist 271,516 9.7%
quote 250,745 9.0%
bullet_list 227,638 8.1%
formula 217,842 7.8%
rule 217,051 7.8%
numbered_list 161,730 5.8%
code 158,996 5.7%
command 104,959 3.8%
image 60,906 2.2%

Repository Provenance

The links above record the source repository state used when this dataset card was generated.

Reproduce This Run

Use the command below to regenerate a dataset with the same publish-time configuration.

uv run main.py generate   --repo-id "junyeong-nero/synthetic-ocr-images-ko-1M"   --lang ko   --size 1000000   --markdown-renderer playwright   --style-profile balanced   --novelty-window 80   --novelty-threshold 0.95   --novelty-max-attempts 4   --formula-source-mode mixed   --formula-dataset-weight 0.45   --formula-random-weight 0.3   --formula-synthetic-weight 0.25   --train-ratio 0.9   --test-ratio 0.1   --seed 42

Details

Repository
junyeong-nero/synthetic-ocr-images-ko-1M
Publisher
Junyeong Song
Task category
Image to text
Tags
ocr, synthetic, markdown
Size category
100K<n<1M
Languages
ko
Revision
5c72a08b6ed3d247a40da564a73679df5b7aceb5
Last updated
2026-10-01

Files

272 files, 114.8 GB in total.

Data270 files · 114.8 GB
Documentation1 file · 3.9 KB
Repository1 file · 2.5 KB
Every file
FileTypeSizeSHA-256
data/test-000000.parquetData89.2 MBe9ab3df4472a
data/test-000001.parquetData85.1 MB023a627fa781
data/test-000002.parquetData86.5 MBacb104feb2f0
data/test-000003.parquetData85.6 MB7c156ff3f758
data/test-000004.parquetData82.4 MB2355dba089be
data/test-000005.parquetData83.8 MBb688551a0dca
data/test-000006.parquetData87.7 MB9e6dc528b093
data/test-000007.parquetData79.3 MBa631f1971b2a
data/test-000008.parquetData81.6 MB18b93c947c8e
data/test-000009.parquetData78.7 MBd4653ed8dde5
data/test-000010.parquetData86.2 MB06eb0532675d
data/test-000011.parquetData88.1 MB4985a648f694
data/test-000012.parquetData82.6 MBe5081dbcabd8
data/test-000013.parquetData87.1 MB7753704166df
data/test-000014.parquetData84.7 MBe94ea3a77c7d
data/test-000015.parquetData85.9 MB6969c77f8852
data/test-000016.parquetData87.3 MBba92aa94319f
data/test-000017.parquetData84.9 MB63075058de47
data/test-000018.parquetData84.8 MBeea2d6ae79ee
data/test-000019.parquetData79.1 MB5985aba5b5ee
data/test-000020.parquetData88.8 MB6093eda5e027
data/test-000021.parquetData87.4 MB90a70d3881c1
data/test-000022.parquetData86.4 MB978c29eaf2de
data/test-000023.parquetData89.2 MB281cce82d2e8
data/test-000024.parquetData83.8 MB0fe42fcb9e2a
data/test-000025.parquetData88.2 MBfd1bfe021a57
data/test-000026.parquetData87.1 MBb59a68b262a8
data/test-000027.parquetData86.2 MBfe997d9dc5c3
data/test-000028.parquetData78.1 MB8a4eb2450b88
data/test-000029.parquetData83.7 MBe4a643bd25de
data/test-000030.parquetData92.4 MBb466b88860ac
data/test-000031.parquetData82.5 MBd8e705bc45e8
data/test-000032.parquetData86.7 MB6993d953fe16
data/test-000033.parquetData85.7 MB04bd42ca4af2
data/test-000034.parquetData86.9 MB02c7675366b8
data/test-000035.parquetData86.0 MBbe9f9a9f877f
data/test-000036.parquetData83.1 MB600bb674684e
data/test-000037.parquetData86.0 MB93c8079b6d2d
data/test-000038.parquetData83.1 MB1e4e4c86efa4
data/test-000039.parquetData88.6 MB1a0569358bdd
data/test-000040.parquetData85.6 MBbe27e2d964bb
data/test-000041.parquetData88.2 MBbb860a7133bd
data/test-000042.parquetData82.0 MBf4acb621098d
data/test-000043.parquetData84.9 MBe917ef8e5fa0
data/test-000044.parquetData88.6 MB82de1b13fed9
data/test-000045.parquetData89.1 MB10bf86f022d5
data/test-000046.parquetData85.4 MBb7e478607339
data/test-000047.parquetData88.2 MB79d9d94d1e45
data/test-000048.parquetData83.1 MB2120422eddf2
data/test-000049.parquetData85.1 MB88f9563d799f
data/test-000050.parquetData81.2 MBb7b5aacffee8
data/test-000051.parquetData83.2 MB5c7309757c29
data/test-000052.parquetData90.2 MBf1ac25b859e8
data/test-000053.parquetData85.3 MB62b84ac5cad9
data/test-000054.parquetData87.4 MB4cda71d3045f
data/test-000055.parquetData83.5 MB54f7ca7848ca
data/test-000056.parquetData84.1 MB114cd36b7c08
data/test-000057.parquetData82.8 MB0995b08508d2
data/test-000058.parquetData87.9 MB4a1d7a788478
data/test-000059.parquetData86.7 MB0dd6e0dd80b7
data/test-000060.parquetData87.4 MBaf529253da17
data/test-000061.parquetData81.6 MBf0acfdeb5993
data/test-000062.parquetData85.5 MBf84a8f6b6f08
data/test-000063.parquetData88.9 MB2cb4200a97eb
data/test-000064.parquetData83.0 MB575b5142b8af
data/test-000065.parquetData95.7 MBfda0337095f1
data/test-000066.parquetData78.2 MBbea8529c213a
data/test-000067.parquetData86.7 MBbac17ee91a50
data/test-000068.parquetData89.3 MB359f4b4dc986
data/test-000069.parquetData83.9 MB31373b775a94
data/test-000070.parquetData87.3 MB8d0ec9fc7820
data/test-000071.parquetData81.3 MBef8ee2fa5bdf
data/test-000072.parquetData88.4 MB5136c0077123
data/test-000073.parquetData82.3 MB8713a69d9b29
data/test-000074.parquetData92.4 MB6f6536748b51
data/test-000075.parquetData83.9 MB9eed1e9ea407
data/test-000076.parquetData78.9 MB46935879c6a0
data/test-000077.parquetData88.7 MB6d5fe68d2518
data/test-000078.parquetData86.6 MBee7c2478580e
data/test-000079.parquetData79.2 MBc0fcf7307431
data/test-000080.parquetData85.9 MB2c9f0496a29d
data/test-000081.parquetData84.1 MBc23a4f58d235
data/test-000082.parquetData90.9 MBfdf6291b5eca
data/test-000083.parquetData77.1 MB6e2e00ebbb9d
data/test-000084.parquetData75.4 MB987b4dbf733f
data/test-000085.parquetData87.9 MBe6f0ad29d0f3
data/test-000086.parquetData84.1 MBbd13c9c6ed5c
data/test-000087.parquetData88.0 MB28c43d5dc0a7
data/test-000088.parquetData83.5 MBd3a1d37652cd
data/test-000089.parquetData86.4 MBafdf400f1fd4
data/test-000090.parquetData86.2 MB7b65f8363de8
data/test-000091.parquetData88.4 MB32c6717d3d2e
data/test-000092.parquetData85.6 MBbbd40f1cf8a2
data/test-000093.parquetData89.7 MB283920f74908
data/test-000094.parquetData91.4 MBef9e613b336c
data/test-000095.parquetData81.3 MBf28975756353
data/test-000096.parquetData81.3 MB22bebcb8ba05
data/test-000097.parquetData85.9 MBf1d852f4f559
data/test-000098.parquetData89.7 MBa1f4574c7a8a
data/test-000099.parquetData80.6 MBd59a4ce6946d
data/test-000100.parquetData81.3 MB73119db3126c
data/test-000101.parquetData89.7 MB35769d90c543
data/test-000102.parquetData74.0 MB2995be8a3609
data/test-000103.parquetData87.9 MBd8b4e91180b4
data/test-000104.parquetData87.0 MB5dee7eff1542
data/test-000105.parquetData85.0 MBd192a9af354c
data/test-000106.parquetData88.3 MB1ba45e07b331
data/test-000107.parquetData84.1 MBe417c8081ee8
data/test-000108.parquetData81.5 MB776a5c66c5b3
data/test-000109.parquetData77.0 MBc5e2fe0eaaf5
data/test-000110.parquetData88.9 MBe22ba167d9bd
data/test-000111.parquetData84.9 MB909df0641db5
data/test-000112.parquetData82.3 MB723f0a253b9e
data/test-000113.parquetData90.5 MB17e238f4330b
data/test-000114.parquetData87.1 MB4f856aed77f1
data/test-000115.parquetData85.4 MB24baea2a549e
data/test-000116.parquetData80.9 MBb6e2e37c1dda
data/test-000117.parquetData80.9 MB0a89837397ca
data/test-000118.parquetData82.5 MB6759d77e85da
data/test-000119.parquetData90.0 MBa19df7cfb26b
data/test-000120.parquetData85.5 MB47217255677c
data/test-000121.parquetData81.2 MB51e425f31f85
data/test-000122.parquetData83.3 MB43be150bfded
data/test-000123.parquetData82.2 MBc13a77bcff0a
data/test-000124.parquetData86.7 MB928fef43e98a
data/test-000125.parquetData85.3 MBcb7e9e4455c1
data/test-000126.parquetData86.3 MBb1f3e4f0f36a
data/test-000127.parquetData94.4 MBc6cbbc619be5
data/test-000128.parquetData93.3 MBcf3664df0ce2
data/test-000129.parquetData84.6 MB6971a1d4d664
data/test-000130.parquetData85.1 MB58555ecb6df4
data/test-000131.parquetData82.9 MBb7414a975981
data/test-000132.parquetData84.3 MBb40b3fb26232
data/test-000133.parquetData83.9 MB43a88abd50de
data/test-000134.parquetData88.9 MBc36378a737f5
data/train-000000.parquetData782.6 MB2eb8ec0bef36
data/train-000001.parquetData759.9 MBe112df5b29a8
data/train-000002.parquetData770.0 MB3c481de9a6d3
data/train-000003.parquetData756.3 MBd000dcda5b09
data/train-000004.parquetData759.5 MBc61eb15c412c
data/train-000005.parquetData770.1 MBb04611b56c80
data/train-000006.parquetData768.8 MB44714f5db9ae
data/train-000007.parquetData781.4 MBbe5d82223746
data/train-000008.parquetData749.3 MB4695aae9867b
data/train-000009.parquetData765.0 MB6144c0cb7f3d
data/train-000010.parquetData771.7 MBdaee37e45fdb
data/train-000011.parquetData769.5 MB1fe4c1f8800e
data/train-000012.parquetData760.9 MB647542ab8580
data/train-000013.parquetData767.6 MBb0b175e479fc
data/train-000014.parquetData770.4 MB1cc6f1ab6aaa
data/train-000015.parquetData765.8 MB9ecd8cb084f3
data/train-000016.parquetData762.9 MB25ccd72167ab
data/train-000017.parquetData799.9 MB71ac08c4f397
data/train-000018.parquetData748.7 MB96757ed8f767
data/train-000019.parquetData777.4 MBa028769899e2
data/train-000020.parquetData758.0 MB08166dd4f288
data/train-000021.parquetData775.0 MB25048853e991
data/train-000022.parquetData754.1 MBcda4e43fcfed
data/train-000023.parquetData754.2 MB367908958134
data/train-000024.parquetData760.5 MBd8b720d7839f
data/train-000025.parquetData752.3 MBd6a39147b2af
data/train-000026.parquetData765.8 MB5f7e116cea1b
data/train-000027.parquetData767.1 MB32c9a53917a7
data/train-000028.parquetData765.9 MB2168e67f785e
data/train-000029.parquetData755.8 MB7098e2cb6c6c
data/train-000030.parquetData773.4 MB497515b10c84
data/train-000031.parquetData742.7 MB137045275d75
data/train-000032.parquetData743.8 MBaa8354370bae
data/train-000033.parquetData761.5 MBe7e0fd1bc1ce
data/train-000034.parquetData762.2 MBec25de05b34b
data/train-000035.parquetData767.4 MB39a2db64a3c6
data/train-000036.parquetData759.9 MB2b080d6fc755
data/train-000037.parquetData762.3 MB043648a91bcc
data/train-000038.parquetData787.6 MB822d71ebf530
data/train-000039.parquetData781.9 MBbd47294fc8b1
data/train-000040.parquetData762.4 MBd98997800003
data/train-000041.parquetData740.9 MB3bad4e17ddb6
data/train-000042.parquetData752.8 MBa1441cd048cc
data/train-000043.parquetData784.7 MBf488251c284b
data/train-000044.parquetData744.8 MB429a0d4a376f
data/train-000045.parquetData780.4 MBa78f4fb9a246
data/train-000046.parquetData777.6 MBe7aa48310348
data/train-000047.parquetData756.2 MB90388911cea4
data/train-000048.parquetData763.5 MBd2baa0c0f274
data/train-000049.parquetData748.8 MBc77ce9e01691
data/train-000050.parquetData736.6 MBcddcb005afa8
data/train-000051.parquetData763.8 MBe288ce3b0add
data/train-000052.parquetData773.8 MB4b819c9c0db0
data/train-000053.parquetData770.8 MB1523bdc48f5b
data/train-000054.parquetData758.6 MB247c47fd82f0
data/train-000055.parquetData771.7 MB03fbab25174b
data/train-000056.parquetData774.3 MB8046235eec28
data/train-000057.parquetData755.6 MB02fa9ce8c1da
data/train-000058.parquetData767.1 MB525d328561c4
data/train-000059.parquetData778.8 MB6399ba6816b5
data/train-000060.parquetData770.7 MBe0ec557f48f5
data/train-000061.parquetData765.6 MB12715be3e4a5
data/train-000062.parquetData756.1 MBd678fb42def7
data/train-000063.parquetData777.7 MB3f0feb432ae4
data/train-000064.parquetData784.7 MB7709f43c6398
data/train-000065.parquetData785.4 MB9cbf34593be6
data/train-000066.parquetData772.6 MB5a6752323c3c
data/train-000067.parquetData759.0 MBa54a5d4b55b2
data/train-000068.parquetData782.2 MBf663e701ae91
data/train-000069.parquetData733.4 MB1c169266184a
data/train-000070.parquetData753.0 MB215b5cd85687
data/train-000071.parquetData740.2 MB642e3a3490d8
data/train-000072.parquetData764.2 MBf5c79269093b
data/train-000073.parquetData763.9 MB316bd65fe8bb
data/train-000074.parquetData771.8 MBcedb21744b44
data/train-000075.parquetData783.0 MB7240e13f2461
data/train-000076.parquetData751.2 MB4b50c63327c4
data/train-000077.parquetData753.1 MBd0a4072554c2
data/train-000078.parquetData805.3 MB3e9bcce30ebc
data/train-000079.parquetData761.9 MB58b905661d5c
data/train-000080.parquetData781.1 MB8c4a2538bd17
data/train-000081.parquetData754.5 MBffd2022eed1e
data/train-000082.parquetData795.2 MB74f6b8fee554
data/train-000083.parquetData755.0 MB74dbc7183500
data/train-000084.parquetData741.7 MB04c78209eeaa
data/train-000085.parquetData754.3 MBd249d618fa60
data/train-000086.parquetData759.0 MBbd5e4091402a
data/train-000087.parquetData762.7 MBc00c30f3ce97
data/train-000088.parquetData746.2 MBa62524ab66e5
data/train-000089.parquetData785.4 MB5b849de634e8
data/train-000090.parquetData765.2 MBfdb1d7d483d7
data/train-000091.parquetData761.1 MB729f8d4d0f2f
data/train-000092.parquetData745.0 MBbfddc8941b65
data/train-000093.parquetData760.7 MBbd65b1365e5a
data/train-000094.parquetData785.2 MB0afb8bd0d448
data/train-000095.parquetData753.9 MB39119c65809b
data/train-000096.parquetData781.5 MB96e75d43507c
data/train-000097.parquetData792.8 MBa5341b1e0f94
data/train-000098.parquetData752.7 MBa3ac943cb32e
data/train-000099.parquetData764.0 MB73aef109f9b5
data/train-000100.parquetData773.0 MB602cf610070e
data/train-000101.parquetData793.1 MB7ae8361f9c41
data/train-000102.parquetData755.7 MB5228f5f3760b
data/train-000103.parquetData757.3 MBc0e5ba912184
data/train-000104.parquetData753.5 MBc90a5a8947dd
data/train-000105.parquetData765.4 MB736c5ecddeb3
data/train-000106.parquetData762.6 MB99326fb6f4dd
data/train-000107.parquetData787.9 MB06dfe0fd8a36
data/train-000108.parquetData753.4 MB2f942c22bb0c
data/train-000109.parquetData769.4 MBf1ade03c0ca4
data/train-000110.parquetData754.5 MBa164fb43f31a
data/train-000111.parquetData744.1 MB32dd8e4a86a7
data/train-000112.parquetData775.4 MBc2fec07a3b6d
data/train-000113.parquetData767.4 MB340a4d5f151c
data/train-000114.parquetData765.8 MBfc212abca989
data/train-000115.parquetData774.3 MBd6acb0f97dea
data/train-000116.parquetData750.3 MB36cdcabeec9c
data/train-000117.parquetData767.6 MB35bb295448c9
data/train-000118.parquetData755.2 MBe4f93ecc5550
data/train-000119.parquetData760.8 MBebe6f48b8cac
data/train-000120.parquetData758.4 MBbcb2320f5975
data/train-000121.parquetData796.2 MB217fb43c0997
data/train-000122.parquetData750.2 MB49f2eb764bbc
data/train-000123.parquetData767.8 MBb5e82da8b093
data/train-000124.parquetData755.2 MB89639911f9f4
data/train-000125.parquetData802.8 MBdc81f893d587
data/train-000126.parquetData740.8 MB22e6c64f49a3
data/train-000127.parquetData752.6 MBf9e08bbda704
data/train-000128.parquetData774.0 MB6d44642e5505
data/train-000129.parquetData756.8 MB4b91757f5a41
data/train-000130.parquetData774.1 MB59bc0e452fe7
data/train-000131.parquetData757.3 MBc8d32ccd4bfd
data/train-000132.parquetData762.4 MB339402c99c56
data/train-000133.parquetData748.4 MB8d2590483530
data/train-000134.parquetData778.1 MB9007324b0b71
README.mdDocumentation3.9 KB—
.gitattributesRepository2.5 KB—

License and Download

License
cc-by-sa-3.0
Access
No access gate
Download from Junyeong Song

Released by Junyeong Song through its official repository on Hugging Face.