Optical Character Recognition2020en

icdar2013

Dataset from Papers With Code

Metrics:accuracy, cer, wer, f1
Legacy BenchmarkLast significant update: Jan 2019

Legacy benchmark from 2013. For current OCR evaluation, use OCRBench v2, ICDAR 2015, or newer benchmarks.

Recommended alternatives:
Current State of the Art

DTrOCR 105M

Unknown

99.4

accuracy

accuracy Progress Over Time

Showing 14 breakthroughs from Jun 2014 to Aug 2023

77.583.589.595.4101.4Jun 2014Mar 2016Jan 2018Nov 2019Sep 2021Aug 2023accuracyDate

Key Milestones

Jun 2014
CHAR

From paper: Synthetic Data and Artificial Neural Networks for Natural Scene Text Recognition

79.5
Jul 2015
CRNN

From paper: An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition

86.7
+9.1%
Mar 2016
RARE

From paper: Robust Scene Text Recognition with Automatic Rectification

88.6
+2.2%
Sep 2016
STAR-Net

From paper: Star-net: A spatial attention residue network for scene text recognition.

89.1
+0.6%
Jun 2018
ASTER

From paper: ASTER: An Attentional Scene Text Recognizer with Flexible Rectification

91.8
+3.0%
Apr 2019
Baek et al.

From paper: What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis

92.3
+0.5%
Oct 2019
SATRN

From paper: On Recognizing Texts of Arbitrary Shapes with 2D Self-Attention

94.1
+2.0%
Mar 2020
SRN

From paper: Towards Accurate Scene Text Recognition with Semantic Reasoning Networks

95.5
+1.5%
Jul 2021
Yet Another Text Recognizer

From paper: Why You Should Try the Real Data for the Scene Text Recognition

96.8
+1.4%
Aug 2021
DPAN

From paper: Look Back Again: Dual Parallel Attention Network for Accurate and Robust Scene Text Recognition

97.7
+0.9%
Nov 2021
MATRN

From paper: Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features

97.9
+0.2%
Sep 2022
MGP-STR

From paper: Multi-Granularity Prediction for Scene Text Recognition

98.5
+0.6%
May 2023
CLIP4STR-L (DataComp-1B)

From paper: CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model

99.0
+0.5%
Aug 2023
DTrOCR 105MCurrent SOTA

From paper: DTrOCR: Decoder-only Transformer for Optical Character Recognition

99.4
+0.4%
Total Improvement
25.0%
Time Span
9y 4m
Breakthroughs
14
Current SOTA
99.4

Top Models Performance Comparison

Top 10 models ranked by accuracy

accuracy1DTrOCR 105M99.4100.0%2CLIP4STR-L (DataComp-1B)99.099.6%3MGP-STR98.599.1%4CLIP4STR-L98.599.1%5CLIP4STR-B*98.398.9%6CCD-ViT-Base(ARD_2.8M)98.398.9%7CCD-ViT-Small(ARD_2.8M)98.398.9%8MATRN97.998.5%9S-GTR97.898.4%10SIGA_T97.898.4%0%25%50%75%100%% of best
Best Score
99.4
Top Model
DTrOCR 105M
Models Compared
10
Score Range
1.6

accuracyPrimary

#ModelScorePaper / CodeDate
1
DTrOCR 105M
99.4Aug 2023
2
CLIP4STR-L (DataComp-1B)
99May 2023
3
MGP-STR
98.5Sep 2022
4
CLIP4STR-L
98.5May 2023
5
CLIP4STR-B*
98.3May 2023
6
CCD-ViT-Base(ARD_2.8M)
98.3Nov 2022
7
CCD-ViT-Small(ARD_2.8M)
98.3Nov 2022
8
MATRN
97.9Nov 2021
9
S-GTR
97.8Dec 2021
10
SIGA_T
97.8Mar 2022
11
DPAN
97.7
Look Back Again: Dual Parallel Attention Network for Accurate and Robust Scene Text RecognitionCode
Aug 2021
12
CDistNet (Ours)
97.67Nov 2021
13
CCD-ViT-Tiny(ARD_2.8M)
97.5Nov 2022
14
SVTR-L (Large)
97.2Apr 2022
15
SVTR-B (Base)
97.1Apr 2022
16
DiffusionSTR
97.1Jun 2023
17
Yet Another Text Recognizer
96.8Jul 2021
18
SVTR-T (Tiny)
96.3Apr 2022
19
SVTR-S (Small)
95.7Apr 2022
20
SRN
95.5Mar 2020
21
RCEED
94.7Jun 2021
22
SATRN
94.1Oct 2019
23
DAN
93.9Dec 2019
24
CSTR
93.2Feb 2021
25
TextScanner
92.9Dec 2019
26
SAFL
92.8Jan 2022
27
SEED
92.8May 2020
28
ViTSTR
92.4May 2021
29
Baek et al.
92.3Apr 2019
30
ASTER
91.8
ASTER: An Attentional Scene Text Recognizer with Flexible RectificationCode
Jun 2018
31
CA-FCN
91.5Sep 2018
32
SAR
91Nov 2018
33
STAR-Net
89.1
Star-net: A spatial attention residue network for scene text recognition.Code
Sep 2016
34
RARE
88.6Mar 2016
35
CRNN
86.7Jul 2015
36
CHAR
79.5Jun 2014

avg-f1

Related Papers29

Self-supervised Character-to-Character Distillation for Text Recognition
Nov 2022Models: CCD-ViT-Base(ARD_2.8M), CCD-ViT-Small(ARD_2.8M), CCD-ViT-Tiny(ARD_2.8M)
SVTR: Scene Text Recognition with a Single Visual Model
Apr 2022Models: SVTR-L (Large), SVTR-B (Base), SVTR-T (Tiny) +1 more

Other Optical Character Recognition Datasets