Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

Text Recognition research benchmarks – Page 2 · SOTA2 Research

Text Recognition

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

ICDAR robust reading 2013 (test)

90.8Accuracy

6

Feb 26, 2026

Full-resolution

33.3Transcription Rate

5

Jun 2, 2026

Latin-script evaluation benchmark

PP-OCRv6_medium

88Accuracy

4

Jun 12, 2026

30-page PG (test)

PG fine-tuned model

1.05CER (%)

4

Mar 11, 2026

Multi-domain Text Recognition Benchmark (Street, Synthetic, Documentation, Handwritten, Car License) (test)

22.43Error Rate (St,Sy,D,H -> C)

4

Feb 26, 2026

In-house-OCR High Resolution

92.32Avg Normalized Edit Distance

4

Feb 27, 2026

In-house-OCR Art Font

92.1Avg Normalized Edit Distance

4

Feb 27, 2026

In-house-OCR Vertical Text

71.28Avg Normalized Edit Distance

4

Feb 27, 2026

In-house-OCR Handwritten

86.29Avg Normalized Edit Distance

4

Feb 27, 2026

OpenScore String Quartets Camera

10.6CER (Title)

3

Jul 8, 2026

Ours (VQA-augmented)

27.26WER

3

Mar 5, 2026

MLT 2019 (test)

66.7Accuracy

3

Feb 26, 2026

Uber-Text (test)

54.86Accuracy

3

Feb 26, 2026

ArT 2019 (test)

58.82Accuracy

3

Feb 26, 2026

95.2Accuracy

1

Feb 26, 2026

—

—Accuracy (small)

0

Feb 26, 2026

Page 2 of 2

Popular datasets

In-house-OCR ICDAR IIIT5K OmniDocBench IIIT

Follow for updates

Join SOTA2 Contact