Datasets:

interfaze-ai
/

ocr-mlt-50m

image image	text string	language string	script string	source_type string	confidence float64
Not supported with pagination yet	Never gonna give you up	en	Latin	signage	0.88
Not supported with pagination yet	永远不会放弃你	zh	CJK	document	0.89
Not supported with pagination yet	決してあなたを諦めない	ja	CJK	scene_text	0.9
Not supported with pagination yet	널 절대 포기하지 않을 거야	ko	CJK	handwritten	0.91
Not supported with pagination yet	لن أتخلى عنك أبداً	ar	Arabic	signage	0.92
Not supported with pagination yet	तुम्हें कभी नहीं छोड़ूंगा	hi	Devanagari	signage	0.93
Not supported with pagination yet	Ich werde dich niemals aufgeben	de	Latin	document	0.94
Not supported with pagination yet	Je ne te laisserai jamais tomber	fr	Latin	scene_text	0.95
Not supported with pagination yet	Nunca te voy a abandonar	es	Latin	handwritten	0.96
Not supported with pagination yet	Nunca vou desistir de você	pt	Latin	signage	0.97
Not supported with pagination yet	Никогда тебя не брошу	ru	Cyrillic	signage	0.98
Not supported with pagination yet	จะไม่มีวันยอมแพ้เธอ	th	Thai	document	0.99
Not supported with pagination yet	Sẽ không bao giờ từ bỏ bạn	vi	Latin	scene_text	0.88
Not supported with pagination yet	Non ti abbandonerò mai	it	Latin	handwritten	0.89
Not supported with pagination yet	Ik zal je nooit opgeven	nl	Latin	signage	0.9
Not supported with pagination yet	Nigdy cię nie porzucę	pl	Latin	signage	0.91
Not supported with pagination yet	Seni asla bırakmayacağım	tr	Latin	document	0.92
Not supported with pagination yet	Jag kommer aldrig ge upp dig	sv	Latin	scene_text	0.93
Not supported with pagination yet	Nikdy tě nevzdám	cs	Latin	handwritten	0.94
Not supported with pagination yet	Nu te voi abandona niciodată	ro	Latin	signage	0.95
Not supported with pagination yet	Jeg vil aldrig give dig op	da	Latin	signage	0.96
Not supported with pagination yet	En koskaan luovu sinusta	fi	Latin	document	0.97
Not supported with pagination yet	Soha nem hagylak el	hu	Latin	scene_text	0.98
Not supported with pagination yet	Δεν θα σε εγκαταλείψω ποτέ	el	Greek	handwritten	0.99
Not supported with pagination yet	Никога няма да те изоставя	bg	Cyrillic	signage	0.88
Not supported with pagination yet	Ніколи тебе не покину	uk	Cyrillic	signage	0.89
Not supported with pagination yet	Nikada te neću ostaviti	hr	Latin	document	0.9
Not supported with pagination yet	Nikdy ťa neopustím	sk	Latin	scene_text	0.91
Not supported with pagination yet	Nikoli te ne bom zapustil	sl	Latin	handwritten	0.92
Not supported with pagination yet	Niekada tavęs neapleisiu	lt	Latin	signage	0.93
Not supported with pagination yet	Es tevi nekad nepametīšu	lv	Latin	signage	0.94
Not supported with pagination yet	Ma ei jäta sind kunagi maha	et	Latin	document	0.95
Not supported with pagination yet	Qatt mhu se nċedilek	mt	Latin	scene_text	0.96
Not supported with pagination yet	Ní thabharfaidh mé suas thú go deo	ga	Latin	handwritten	0.97
Not supported with pagination yet	Saya tidak akan pernah menyerah padamu	ms	Latin	signage	0.98
Not supported with pagination yet	Saya tidak akan pernah menyerahkanmu	id	Latin	signage	0.99
Not supported with pagination yet	Hindi kita kailanman iiwan	tl	Latin	document	0.88
Not supported with pagination yet	Sitakuacha kamwe	sw	Latin	scene_text	0.89
Not supported with pagination yet	በፍጹም አልተውህም	am	Ethiopic	handwritten	0.9
Not supported with pagination yet	তোমাকে কখনো ছেড়ে দেব না	bn	Bengali	signage	0.91
Not supported with pagination yet	உன்னை ஒருபோதும் கைவிடமாட்டேன்	ta	Tamil	signage	0.92
Not supported with pagination yet	నిన్ను ఎప్పటికీ వదలను	te	Telugu	document	0.93
Not supported with pagination yet	ನಿನ್ನನ್ನು ಎಂದಿಗೂ ಬಿಡುವುದಿಲ್ಲ	kn	Kannada	scene_text	0.94
Not supported with pagination yet	നിന്നെ ഒരിക്കലും കൈവിടില്ല	ml	Malayalam	handwritten	0.95
Not supported with pagination yet	તને ક્યારેય છોડીશ નહીં	gu	Gujarati	signage	0.96
Not supported with pagination yet	तुला कधीच सोडणार नाही	mr	Devanagari	signage	0.97
Not supported with pagination yet	ਤੈਨੂੰ ਕਦੇ ਨਹੀਂ ਛੱਡਾਂਗਾ	pa	Gurmukhi	document	0.98
Not supported with pagination yet	میں تمہیں کبھی نہیں چھوڑوں گا	ur	Arabic	scene_text	0.99
Not supported with pagination yet	तिमीलाई कहिल्यै छोड्ने छैन	ne	Devanagari	handwritten	0.88
Not supported with pagination yet	මම ඔබව කවදාවත් අත් නොහරිමි	si	Sinhala	signage	0.89

OCR-MLT-50M: Multilingual OCR Corpus

A large-scale multilingual OCR dataset spanning 50 languages and 50.2 million image-text pairs.
Designed for training and evaluating robust multilingual text recognition systems across diverse scripts and domains.

🔥 News

[2025-11-15] OCR-MLT-50M is now available on Hugging Face! Download here
[2025-10-28] Our paper is accepted at CVPR 2025! Camera-ready version
[2025-09-10] Released v2 model weights with improved CJK performance. Model card
[2025-08-01] Pre-trained checkpoints available for all 50 languages. Download

Overview

Stat	Value
Total samples	50,217,843
Languages	50
Scripts	14 (Latin, CJK, Arabic, Devanagari, Cyrillic, ...)
Source types	Scene text, documents, handwritten, receipts, signage
Avg. image resolution	384 x 128
Storage (compressed)	~2.3 TB

Language Distribution

Click to view the full interactive breakdown by language and script family

Sample Visualizations



Scene text (EN)	Document (JA)	Handwritten (AR)

Data Collection Pipeline

Samples were collected from three primary sources:

Synthetic rendering — text rendered onto natural backgrounds using 2,400+ fonts per script
Web-crawled scene text — filtered and deduplicated from Common Crawl with PaddleOCR pseudo-labels
Scanned documents — partnerships with national libraries and digitization initiatives

All pseudo-labels were verified using a multi-model consensus approach (TrOCR + PaddleOCR + EasyOCR), retaining only samples with ≥2/3 agreement. Full methodology in our technical report.

Quick Start

from datasets import load_dataset

# Load a specific language split
ds = load_dataset("interfaze-ai/ocr-mlt-50m", "en", split="train", streaming=True)

for sample in ds:
    print(sample["text"], sample["language"])
    break

Benchmarks

Models fine-tuned on OCR-MLT-50M vs. existing public corpora:

Model	MLT-2019 (F1)	IC15 (Acc)	CUTE80 (Acc)	Details
TrOCR-large + Ours	87.3	96.1	94.7	Config & Weights
PARSeq + Ours	88.1	96.8	95.2	Config & Weights
CLIP4STR + Ours	89.6	97.2	96.0	Config & Weights
Baseline (MJSynth+ST)	79.4	94.2	87.8	—

Full evaluation scripts and configs: GitHub

Shards

Data is split into per-language shards. See the file listing for the full manifest.

Citation

@inproceedings{kumar2025ocrmlt,
  title={OCR-MLT-50M: Scaling Multilingual Text Recognition with Synthetic-Real Hybrid Corpora},
  author={Kumar, Arjun and Nakamura, Yui and Al-Rashid, Fatima and M{\"u}ller, Jonas},
  booktitle={Proceedings of CVPR 2025},
  year={2025},
  pages={11234--11245}
}

License

Apache 2.0 — see LICENSE for details.

Downloads last month: 13

Size of downloaded dataset files:

9.9 kB

Size of the auto-converted Parquet files:

9.9 kB

Number of rows: