Codesota · Computer Vision · Scene Text Recognition · svtTasks/Computer Vision/Scene Text Recognition
Scene Text Recognition · benchmark dataset · 2020 · EN

svt.

Dataset from Papers With Code

Submit a result ↵
§ 01 · Leaderboard

Best published scores.

40 results indexed across 1 metric. Shaded row marks current SOTA; ties broken by submission date.

Primary
accuracy · higher is better
accuracy· primary
40 rows
#ModelOrgSubmittedPaper / codeaccuracy
01CLIP4STR-H (DFN-5B)—May 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code99.10
02DTrOCR 105M—Aug 2023DTrOCR: Decoder-only Transformer for Optical Character R… · code98.90
03MGP-STR—Sep 2022Multi-Granularity Prediction for Scene Text Recognition · code98.60
04CLIP4STR-L (DataComp-1B)—May 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code98.60
05CPPD—Jul 2023Context Perception Parallel Decoder for Scene Text Recog… · code98.50
06CLIP4STR-L—May 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code98.50
07CLIP4STR-BResearchMay 2023CLIP4STR: A Simple Baseline for Scene Text Recognition w… · code98.30
08PARSeqOSSResearchJul 2022Scene Text Recognition with Permuted Autoregressive Sequ…97.84
09CCD-ViT-Base(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code97.80
10CCD-ViT-Small(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code96.40
11TrOCR-large 558M—Sep 2021TrOCR: Transformer-based Optical Character Recognition w…96.10
12CCD-ViT-Tiny(ARD_2.8M)—Nov 2022Self-supervised Character-to-Character Distillation for … · code96
13S-GTR—Dec 2021Visual Semantics Allow for Textual Reasoning Better in S… · code95.80
14TrOCR-base 334M—Sep 2021TrOCR: Transformer-based Optical Character Recognition w…95.20
15SIGA_T—Mar 2022Self-supervised Implicit Glyph Attention for Text Recogn… · code95.10
16MATRNResearchNov 2021Multi-modal Text Recognition Networks: Interactive Enhan… · code95
17Yet Another Text Recognizer—Jul 2021Why You Should Try the Real Data for the Scene Text Reco… · code94.70
18NRTR+TPS++—May 2023TPS++: Attention-Enhanced Thin-Plate Spline for Scene Te… · code94.60
19DPAN—Aug 2021papers-with-code · code93.90
20CDistNet (Ours)—Nov 2021CDistNet: Perceiving Multi-Domain Character Distance for… · code93.82
21DiffusionSTR—Jun 2023DiffusionSTR: Diffusion Model for Scene Text Recognition93.60
22ABINet-LVOSSFang et al.Mar 2021Read Like Humans: Autonomous, Bidirectional and Iterativ…93.40
23RCEED—Jun 2021Representation and Correlation Enhanced Encoder-Decoder … · code91.80
24SRN—Mar 2020Towards Accurate Scene Text Recognition with Semantic Re… · code91.50
25SATRN—Oct 2019On Recognizing Texts of Arbitrary Shapes with 2D Self-At… · code91.30
26CSTR—Feb 2021Revisiting Classification Perspective on Scene Text Reco… · code90.60
27TextScanner—Dec 2019TextScanner: Reading Characters in Order for Robust Scen…90.10
28SEED—May 2020SEED: Semantics Enhanced Encoder-Decoder Framework for S… · code89.60
29ASTER—Jun 2018papers-with-code · code89.50
30DAN—Dec 2019Decoupled Attention Network for Text Recognition · code89.20
31SAFL—Jan 2022SAFL: A Self-Attention Scene Text Recognizer with Focal … · code88.60
32ViTSTR—May 2021Vision Transformer for Fast and Efficient Scene Text Rec… · code87.70
33Baek et al.—Apr 2019What Is Wrong With Scene Text Recognition Model Comparis… · code87.50
34CA-FCN—Sep 2018Scene Text Recognition from Two-Dimensional Perspective86.40
35SAR—Nov 2018Show, Attend and Read: A Simple and Strong Baseline for … · code84.50
36STAR-Net—Sep 2016papers-with-code · code83.60
37RARE—Mar 2016Robust Scene Text Recognition with Automatic Rectificati… · code81.90
38CRNN—Jul 2015An End-to-End Trainable Neural Network for Image-based S… · code80.80
39CRNN—Jul 2015An End-to-End Trainable Neural Network for Image-based S…80.80
40CHAR—Jun 2014Synthetic Data and Artificial Neural Networks for Natura… · code68
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

13 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, 2014CHAR68
  2. Jul 21, 2015CRNN80.80
  3. Mar 12, 2016RARE81.90
  4. Sep 20, 2016STAR-Net83.60
  5. Jun 25, 2018ASTER89.50
  6. Oct 10, 2019SATRN91.30
  7. Mar 27, 2020SRN91.50
  8. Mar 6, 2021ABINet-LVFang et al.93.40
  9. Jul 29, 2021Yet Another Text Recognizer94.70
  10. Sep 21, 2021TrOCR-large 558M96.10
  11. Jul 14, 2022PARSeqResearch97.84
  12. Sep 8, 2022MGP-STR98.60
  13. May 23, 2023CLIP4STR-H (DFN-5B)99.10
Fig 3 · SOTA-setting models only. 13 entries span Jun 2014 → May 2023.
§ 04 · Literature

31 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