Codesota · Benchmark · icdar2015Home/Leaderboards/Vision & Documents/Document OCR/icdar2015
Unknown

icdar2015.

icdar2015 is a state-of-the-art machine learning benchmark indexed on Codesota. This page tracks published model results, top scores per metric, and the SOTA timeline for icdar2015.

Paper ↗Leaderboard ↓
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

Accuracy

Accuracy is the reported evaluation metric for icdar2015. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for Accuracyverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01JSTR
Synthetic training data variant. Table 1 in arxiv:2404.05967. IC15=98.7%. Standard lowercase alphanumeric eval.
verified98.72024Paper ↗Looks wrong?
02TextBlockV2 (GPT-2)
GPT-2 LM decoder variant. Table I in arxiv:2403.10047. IC15=97.7%. Standard lowercase alphanumeric eval.
verified97.72024Paper ↗Looks wrong?
03DTrOCR 105M
From paper: DTrOCR: Decoder-only Transformer for Optical Character Recognition
verified93.52023Paper ↗Code ↗Looks wrong?
04CPPD
From paper: Context Perception Parallel Decoder for Scene Text Recognition
verified91.72023Paper ↗Code ↗Looks wrong?
05CLIP4STR-L (DataComp-1B)
From paper: CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model
verified91.42023Paper ↗Code ↗Looks wrong?
06MGP-STR
From paper: Multi-Granularity Prediction for Scene Text Recognition
verified90.92022Paper ↗Code ↗Looks wrong?
07CLIP4STR-L
From paper: An Empirical Study of Scaling Law for OCR
verified90.82023Paper ↗Code ↗Source ↗Looks wrong?
08CLIP4STR-B
From paper: CLIP4STR: A Simple Baseline for Scene Text Recognition with Pre-trained Vision-Language Model
verified90.62023Paper ↗Code ↗Looks wrong?
09OTSNet
Unified Observation-Thinking-Spelling network. Table 1 in arxiv:2511.08133. IC15=90.2%. Standard lowercase alphanumeric eval.
verified90.22025Paper ↗Looks wrong?
10IGTR-AR
Auto-Regressive variant with real-world training data. Table X in arxiv:2401.17851. IC15=89.8%. Standard lowercase alphanumeric eval.
verified89.82024Paper ↗Looks wrong?
11SIGA_S
From paper: Self-supervised Implicit Glyph Attention for Text Recognition
verified87.62022Paper ↗Code ↗Looks wrong?
12S-GTR
From paper: Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition
verified87.32021Paper ↗Code ↗Looks wrong?
13MATRN
From paper: Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features
verified86.62021Paper ↗Code ↗Looks wrong?
14CDistNet (Ours)
From paper: CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition
verified86.252021Paper ↗Code ↗Looks wrong?
15DiffusionSTR
From paper: DiffusionSTR: Diffusion Model for Scene Text Recognition
verified862023Paper ↗Looks wrong?
16DPAN
From paper: Look Back Again: Dual Parallel Attention Network for Accurate and Robust Scene Text Recognition
verified85.52021Paper ↗Code ↗Looks wrong?
17RCEED
From paper: Representation and Correlation Enhanced Encoder-Decoder Framework for Scene Text Recognition
verified82.22021Paper ↗Code ↗Looks wrong?
18CSTR
From paper: Revisiting Classification Perspective on Scene Text Recognition
verified81.62021Paper ↗Code ↗Looks wrong?
19Yet Another Text Recognizer
From paper: Why You Should Try the Real Data for the Scene Text Recognition
verified80.22021Paper ↗Code ↗Looks wrong?
20SEED
From paper: SEED: Semantics Enhanced Encoder-Decoder Framework for Scene Text Recognition
verified802020Paper ↗Code ↗Looks wrong?
21TextScanner
From paper: TextScanner: Reading Characters in Order for Robust Scene Text Recognition
verified79.42019Paper ↗Looks wrong?
22SATRN
From paper: On Recognizing Texts of Arbitrary Shapes with 2D Self-Attention
verified792019Paper ↗Code ↗Looks wrong?
23SAFL
From paper: SAFL: A Self-Attention Scene Text Recognizer with Focal Loss
verified77.52022Paper ↗Code ↗Looks wrong?
24ASTER
From paper: ASTER: An Attentional Scene Text Recognizer with Flexible Rectification
verified76.12018Paper ↗Code ↗Looks wrong?
25DAN
From paper: Decoupled Attention Network for Text Recognition
verified74.52019Paper ↗Code ↗Looks wrong?
26AON
From paper: AON: Towards Arbitrarily-Oriented Text Recognition
verified732017Paper ↗Code ↗Looks wrong?
27ViTSTR
From paper: Vision Transformer for Fast and Efficient Scene Text Recognition
verified72.62021Paper ↗Code ↗Looks wrong?
28Baek et al.
From paper: What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis
verified71.82019Paper ↗Code ↗Looks wrong?
29SAR
From paper: Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition
verified69.22018Paper ↗Code ↗Looks wrong?

F Measure

F Measure is the reported evaluation metric for icdar2015. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for F Measureverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01DAL
From paper: Dynamic Anchor Learning for Arbitrary-Oriented Object Detection
verified82.42020Paper ↗Code ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Document OCR