Codesota · Benchmark · icdar-2013Home/Leaderboards/Vision & Documents/Scene Text Detection/icdar-2013
Unknown

icdar-2013.

icdar-2013 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 icdar-2013.

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 icdar-2013. 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
IC13, synthetic training data (MJ+ST). Best synth result. IntelliSys 2024. JSTR Table 1.
verified99.22024Paper ↗Looks wrong?
02CLIP4STR-L (RBU 6.5M)
IC13_1015 split. Trained on RBU 6.5M real data. IEEE TIP Dec 2024. CLIP4STR Table III.
verified992023Paper ↗Looks wrong?
03CLIP4STR-H (DFN-5B)
IC13_1015 split. ViT-H/14 pre-trained on DFN-5B. IEEE TIP Dec 2024. CLIP4STR Table III.
verified98.92023Paper ↗Looks wrong?
04DTrOCR
IC13, synthetic training data (MJ+ST). WACV 2024. Verified via JSTR Table 1 (arxiv:2404.05967).
verified98.82023Paper ↗Looks wrong?
05SVTRv2-B
IC13_1015 split. SVTRv2-B (Base). CTC-based. ICCV 2025. Table 3. Best CTC result.
verified98.72024Paper ↗Looks wrong?
06LISTER
IC13_1015 split, lowercase alphanum eval. ICCV 2023. Verified in SVTRv2 Table 3.
verified98.62023Paper ↗Looks wrong?
07SVTRv2-S
IC13_1015 split. SVTRv2-S (Small). CTC-based. ICCV 2025. Table 3.
verified98.52024Paper ↗Looks wrong?
08TrOCR-large 558M
TrOCR-large, Syn+Benchmark training. Table 6. AAAI 2023.
verified98.42021Paper ↗Looks wrong?
09TrOCR-base 334M
TrOCR-base, Syn+Benchmark training. Table 6. AAAI 2023.
verified98.42021Paper ↗Looks wrong?
10CPPD
IC13_1015 split, lowercase alphanum eval. Verified in SVTRv2 Table 3.
verified98.22023Paper ↗Looks wrong?
11MAERec
IC13_1015 split, lowercase alphanum eval. MAERec ViT-B. Verified in SVTRv2 Table 3.
verified98.22023Paper ↗Looks wrong?
12PARSeq
IC13_1015 split, lowercase alphanum eval. ECCV 2022.
verified98.132022Paper ↗Looks wrong?
13SVTRv2-T
IC13_1015 split. SVTRv2-T (Tiny). CTC-based. ICCV 2025. Table 3.
verified982024Paper ↗Looks wrong?
14ABINet-LV
ABINet Language-Vision variant. CVPR 2021.
verified972021Paper ↗Looks wrong?
15CRNN
Lexicon-free. Table 2. TPAMI 2017.
verified86.72015Paper ↗Looks wrong?

Precision

Precision is the reported evaluation metric for icdar-2013. 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 Precisionverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01CRAFT
From paper: Character Region Awareness for Text Detection
verified97.42019Paper ↗Code ↗Looks wrong?
02TextFuseNet (ResNeXt-101)
From paper: TextFuseNet: Scene Text Detection with Richer Fused Features
verified97.272020Paper ↗Code ↗Looks wrong?
03Mask TextSpotter
From paper: Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes
verified952018Paper ↗Code ↗Looks wrong?
04SPCNET
From paper: Scene Text Detection with Supervised Pyramid Context Network
verified93.82018Paper ↗Code ↗Looks wrong?
05WordSup (VGG16-synth-icdar)
From paper: WordSup: Exploiting Word Annotations for Character based Text Detection
verified93.342017Paper ↗Looks wrong?
06Gupta et al.
From paper: Synthetic Data for Text Localisation in Natural Images
verified922016Paper ↗Code ↗Looks wrong?
07Corner Localization (multi-scale)
From paper: Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation
verified922018Paper ↗Code ↗Looks wrong?
08Corner-based Region Proposals
From paper: Detecting Multi-Oriented Text with Corner-based Region Proposals
verified91.92018Paper ↗Code ↗Looks wrong?
09TextBoxes++_MS
From paper: TextBoxes++: A Single-Shot Oriented Scene Text Detector
verified912018Paper ↗Code ↗Looks wrong?
10PixelLink+VGG16 2s MS
From paper: PixelLink: Detecting Scene Text via Instance Segmentation
verified88.62018Paper ↗Code ↗Looks wrong?
11Jaderberg et al.
From paper: Reading Text in the Wild with Convolutional Neural Networks
verified88.52014Paper ↗Looks wrong?
12SSTD
From paper: Single Shot Text Detector with Regional Attention
verified882017Paper ↗Code ↗Looks wrong?
13SegLink
From paper: Detecting Oriented Text in Natural Images by Linking Segments
verified87.72017Paper ↗Code ↗Looks wrong?
14Neumann et al. *
From paper: Efficient Scene Text Localization and Recognition with Local Character Refinement
verified81.82015Paper ↗Looks wrong?

H Mean

H Mean is the reported evaluation metric for icdar-2013. 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 H Meanverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01CRAFT
From paper: Character Region Awareness for Text Detection
verified95.22019Paper ↗Code ↗Looks wrong?

F Measure

F Measure is the reported evaluation metric for icdar-2013. 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
01TextFuseNet (ResNeXt-101)
From paper: TextFuseNet: Scene Text Detection with Richer Fused Features
verified94.612020Paper ↗Code ↗Looks wrong?
02SPCNET
From paper: Scene Text Detection with Supervised Pyramid Context Network
verified92.12018Paper ↗Code ↗Looks wrong?
03Mask TextSpotter
From paper: Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes
verified91.72018Paper ↗Code ↗Looks wrong?
04WordSup (VGG16-synth-icdar)
From paper: WordSup: Exploiting Word Annotations for Character based Text Detection
verified90.342017Paper ↗Looks wrong?
05STN-OCR
From paper: STN-OCR: A single Neural Network for Text Detection and Text Recognition
verified90.32017Paper ↗Code ↗Looks wrong?
06PixelLink+VGG16 2s MS
From paper: PixelLink: Detecting Scene Text via Instance Segmentation
verified88.12018Paper ↗Code ↗Looks wrong?
07Corner Localization (multi-scale)
From paper: Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation
verified882018Paper ↗Code ↗Looks wrong?
08TextBoxes++_MS
From paper: TextBoxes++: A Single-Shot Oriented Scene Text Detector
verified882018Paper ↗Code ↗Looks wrong?
09Corner-based Region Proposals
From paper: Detecting Multi-Oriented Text with Corner-based Region Proposals
verified87.62018Paper ↗Code ↗Looks wrong?
10SSTD
From paper: Single Shot Text Detector with Regional Attention
verified872017Paper ↗Code ↗Looks wrong?
11SegLink
From paper: Detecting Oriented Text in Natural Images by Linking Segments
verified85.32017Paper ↗Code ↗Looks wrong?
12Gupta et al.
From paper: Synthetic Data for Text Localisation in Natural Images
verified832016Paper ↗Code ↗Looks wrong?
13USM (COCO TS + ICDAR–2013)
From paper: Unsharp Masking Layer: Injecting Prior Knowledge in Convolutional Networks for Image Classification
verified80.42019Paper ↗Code ↗Looks wrong?
14Neumann et al. *
From paper: Efficient Scene Text Localization and Recognition with Local Character Refinement
verified77.12015Paper ↗Looks wrong?
15Jaderberg et al.
From paper: Reading Text in the Wild with Convolutional Neural Networks
verified76.82014Paper ↗Looks wrong?

Recall

Recall is the reported evaluation metric for icdar-2013. 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 Recallverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01CRAFT
From paper: Character Region Awareness for Text Detection
verified93.12019Paper ↗Code ↗Looks wrong?
02TextFuseNet (ResNeXt-101)
From paper: TextFuseNet: Scene Text Detection with Richer Fused Features
verified92.092020Paper ↗Code ↗Looks wrong?
03SPCNET
From paper: Scene Text Detection with Supervised Pyramid Context Network
verified90.52018Paper ↗Code ↗Looks wrong?
04Mask TextSpotter
From paper: Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes
verified88.62018Paper ↗Code ↗Looks wrong?
05WordSup (VGG16-synth-icdar)
From paper: WordSup: Exploiting Word Annotations for Character based Text Detection
verified87.532017Paper ↗Looks wrong?
06PixelLink+VGG16 2s MS
From paper: PixelLink: Detecting Scene Text via Instance Segmentation
verified87.52018Paper ↗Code ↗Looks wrong?
07SSTD
From paper: Single Shot Text Detector with Regional Attention
verified862017Paper ↗Code ↗Looks wrong?
08Corner Localization (multi-scale)
From paper: Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation
verified84.42018Paper ↗Code ↗Looks wrong?
09TextBoxes++_MS
From paper: TextBoxes++: A Single-Shot Oriented Scene Text Detector
verified842018Paper ↗Code ↗Looks wrong?
10Corner-based Region Proposals
From paper: Detecting Multi-Oriented Text with Corner-based Region Proposals
verified83.92018Paper ↗Code ↗Looks wrong?
11SegLink
From paper: Detecting Oriented Text in Natural Images by Linking Segments
verified832017Paper ↗Code ↗Looks wrong?
12Gupta et al.
From paper: Synthetic Data for Text Localisation in Natural Images
verified75.52016Paper ↗Code ↗Looks wrong?
13Neumann et al. *
From paper: Efficient Scene Text Localization and Recognition with Local Character Refinement
verified72.42015Paper ↗Looks wrong?
14Jaderberg et al.
From paper: Reading Text in the Wild with Convolutional Neural Networks
verified67.82014Paper ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Scene Text Detection
icdar-2013 Leaderboard | CodeSOTA | CodeSOTA