Codesota · Computer Vision · Optical Character Recognition · icdar2013Tasks/Computer Vision/Optical Character Recognition
Optical Character Recognition · benchmark dataset · 2020 · EN

icdar2013.

Dataset from Papers With Code

Legacy benchmark· last significant update Jan 2019

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

Submit a result ↵
§ 01 · Leaderboard

Best published scores.

39 results indexed across 2 metrics. Shaded row marks current SOTA; ties broken by submission date.

Primary
accuracy · higher is better
All metrics
accuracy, avg-f1
accuracy· primary
36 rows
#ModelOrgSubmittedPaper / codeaccuracy
01DTrOCR 105M—Aug 2023DTrOCR: Decoder-only Transformer for Optical Character R… · code99.40
02CLIP4STR-L (DataComp-1B)—May 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code99
03CLIP4STR-L—May 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code98.50
04MGP-STR—Sep 2022Multi-Granularity Prediction for Scene Text Recognition · code98.50
05CLIP4STR-BResearchMay 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code98.30
06CCD-ViT-Small(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code98.30
07CCD-ViT-Base(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code98.30
08MATRNResearchNov 2021Multi-modal Text Recognition Networks: Interactive Enhan… · code97.90
09SIGA_T—Mar 2022Self-supervised Implicit Glyph Attention for Text Recogn… · code97.80
10S-GTR—Dec 2021Visual Semantics Allow for Textual Reasoning Better in S… · code97.80
11DPAN—Aug 2021papers-with-code · code97.70
12CDistNet (Ours)—Nov 2021CDistNet: Perceiving Multi-Domain Character Distance for… · code97.67
13CCD-ViT-Tiny(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code97.50
14SVTR-L (Large)—Apr 2022SVTR: Scene Text Recognition with a Single Visual Model · code97.20
15SVTR-B (Base)—Apr 2022SVTR: Scene Text Recognition with a Single Visual Model · code97.10
16DiffusionSTR—Jun 2023DiffusionSTR: Diffusion Model for Scene Text Recognition97.10
17Yet Another Text Recognizer—Jul 2021Why You Should Try the Real Data for the Scene Text Reco… · code96.80
18SVTR-T (Tiny)—Apr 2022SVTR: Scene Text Recognition with a Single Visual Model · code96.30
19SVTR-S (Small)—Apr 2022SVTR: Scene Text Recognition with a Single Visual Model · code95.70
20SRN—Mar 2020Towards Accurate Scene Text Recognition with Semantic Re… · code95.50
21RCEED—Jun 2021Representation and Correlation Enhanced Encoder-Decoder … · code94.70
22SATRN—Oct 2019On Recognizing Texts of Arbitrary Shapes with 2D Self-At… · code94.10
23DAN—Dec 2019Decoupled Attention Network for Text Recognition · code93.90
24CSTR—Feb 2021Revisiting Classification Perspective on Scene Text Reco… · code93.20
25TextScanner—Dec 2019TextScanner: Reading Characters in Order for Robust Scen…92.90
26SAFL—Jan 2022SAFL: A Self-Attention Scene Text Recognizer with Focal … · code92.80
27SEED—May 2020SEED: Semantics Enhanced Encoder-Decoder Framework for S… · code92.80
28ViTSTR—May 2021Vision Transformer for Fast and Efficient Scene Text Rec… · code92.40
29Baek et al.—Apr 2019What Is Wrong With Scene Text Recognition Model Comparis… · code92.30
30ASTER—Jun 2018papers-with-code · code91.80
31CA-FCN—Sep 2018Scene Text Recognition from Two-Dimensional Perspective91.50
32SAR—Nov 2018Show, Attend and Read: A Simple and Strong Baseline for … · code91
33STAR-Net—Sep 2016papers-with-code · code89.10
34RARE—Mar 2016Robust Scene Text Recognition with Automatic Rectificati… · code88.60
35CRNN—Jul 2015An End-to-End Trainable Neural Network for Image-based S… · code86.70
36CHAR—Jun 2014Synthetic Data and Artificial Neural Networks for Natura… · code79.50
avg-f1
3 rows
#ModelOrgSubmittedPaper / codeavg-f1
01CDeCNet—Aug 2020CDeC-Net: Composite Deformable Cascade Network for Table… · code1.00
02cascadetabnet—Apr 2020CascadeTabNet: An approach for end to end table detectio… · code1.00
03TableNet—Jan 2020TableNet: Deep Learning model for end-to-end Table detec… · code0.966
Fig 2 · Rows sorted by score within each metric. Shaded row marks SOTA. Dates reflect model or paper release where available, otherwise the date Codesota accessed the source.
§ 03 · Progress

14 steps
of state of the art.

Each row below marks a model that broke the previous record on accuracy. Intermediate submissions are kept in the leaderboard above; only SOTA-setting entries are re-listed here.

Higher scores win. Each subsequent entry improved upon the previous best.

SOTA line · accuracy
  1. Jun 9, 2014CHAR79.50
  2. Jul 21, 2015CRNN86.70
  3. Mar 12, 2016RARE88.60
  4. Sep 20, 2016STAR-Net89.10
  5. Jun 25, 2018ASTER91.80
  6. Apr 3, 2019Baek et al.92.30
  7. Oct 10, 2019SATRN94.10
  8. Mar 27, 2020SRN95.50
  9. Jul 29, 2021Yet Another Text Recognizer96.80
  10. Aug 1, 2021DPAN97.70
  11. Nov 30, 2021MATRNResearch97.90
  12. Sep 8, 2022MGP-STR98.50
  13. May 23, 2023CLIP4STR-L (DataComp-1B)99
  14. Aug 30, 2023DTrOCR 105M99.40
Fig 3 · SOTA-setting models only. 14 entries span Jun 2014 → Aug 2023.
§ 04 · Literature

29 papers
tied to this benchmark.

Every paper below corresponds to at least one row in the leaderboard above. Click through for the arXiv preprint and, when available, the reference implementation.

§ 06 · Contribute

Have a score that beats
this table?

Submit a checkpoint and a reproduction script. We will run it, publish the score, and — if it takes the top — annotate the step on the progress chart with your name.

Submit a result ↵Read submission guide
What a submission needs
  • 01A public checkpoint or API endpoint
  • 02A reproduction script with frozen commit + seed
  • 03Declared evaluation environment (Python, deps)
  • 04One row per metric declared by this dataset
  • 05A contact so we can follow up on discrepancies