Scene Text Recognition on SVTP (test)
98.3Word AccuracyCSD-B
Evaluation Results
| Method | Links | |
|---|---|---|
| CSD-BData=RBU, Params=104.9 M2025.03 | 98.3 | |
| CLIP4STR-L*Training data=REBU-Syn2023.12 | 98.13 | |
| CSD-DData=Real, Params=110.2 M2025.03 | 98.1 | |
| CLIP4STR-LData=RBU, Params=446 M2025.03 | 98.1 | |
| CSD-DData=RBU, Params=110.2 M2025.03 | 98.1 | |
| CLIP4STR-BData=RBU, Params=158 M2025.03 | 98 | |
| CLIP4STR-HData=RBU, Params=1 B2025.03 | 98 | |
| CLIP4STR-B*Training data=REBU-Syn2023.12 | 97.83 | |
| CSD-BData=Real, Params=104.9 M2025.03 | 97.8 | |
| CSD-SData=RBU, Params=40.8 M2025.03 | 97.8 | |
| CSD-SData=Real, Params=40.8 M2025.03 | 97.5 | |
| CLIP4STR-LData=Real, Params=446 M2025.03 | 97.4 | |
| CLIP4STR-LTraining data=Real2023.12 | 97.36 | |
| CLIP4STR-BData=Real, Params=158 M2025.03 | 97.2 | |
| CLIP4STR-BTraining data=Real2023.12 | 96.74 | |
| PARSeqData=Real, Params=22.5 M2025.03 | 95.7 | |
| PARSeqATraining data=Real, Reproduced=true2023.12 | 94.73 | |
| MaskOCR(ViT-L)Training data=MJ+ST2023.12 | 94.6 | |
| MAERec-BTraining data=Union14M-L2023.12 | 94.4 | |
| ABINetTraining data=Real2023.12 | 94.1 | |
| ABINETData=Real, Params=23.5 M2025.03 | 93.9 | |
| TRBAData=Real, Params=49.6 M2025.03 | 93.7 | |
| TrOCRLargeTraining data=MJ+ST+B2023.12 | 93 | |
| DiG-ViT-BTraining data=Real2023.12 | 92.9 | |
| ViTSTR-STraining data=Real2023.12 | 91.5 | |
| VITSTR-SData=Real, Params=21.7 M2025.03 | 91.4 | |
| MGP-STR FuseTraining Data=MJ and ST, Lexicon=None2022.09 | 91.01 | |
| DiG-ViT-BTraining data=MJ+ST2023.12 | 91 | |
| MATRN2021.11 | 90.6 | |
| MATRNTraining data=MJ+ST2023.12 | 90.6 | |
| MaskOCR(ViT-B)Training data=MJ+ST2023.12 | 89.9 | |
| ABINet-LV+TPS++Training Data=90K+ST, Params=37.2M, Time=41.5ms2023.05 | 89.6 | |
| MGP-STR VisionTraining Data=MJ and ST, Lexicon=None2022.09 | 89.46 | |
| ABINetYear=20212021.11 | 89.3 | |
| ABINet (reproduced)Implementation=reproduced2021.11 | 89.3 | |
| ABINetTraining Data=MJ and ST, Lexicon=None2022.09 | 89.3 | |
| ABINet-LVTraining Data=90K+ST, Params=36.7M, Time=37.2ms2023.05 | 89.3 | |
| ABINetYear=2021, Training Data=90K+ST2021.11 | 89.3 | |
| TRBATraining data=MJ+ST2023.12 | 89.3 | |
| ABINetTraining data=MJ+ST2023.12 | 89.3 | |
| ABINet + SeTaPruned %=28.1%2025.03 | 89.3 | |
| NRTR+TPS++Training Data=90K+ST, Params=35.5M, Time=218ms2023.05 | 89 | |
| PARSeqTraining Data=90K+ST, Params=23.8M, Time=11.8ms2023.05 | 88.9 | |
| PARSeqATraining data=MJ+ST2023.12 | 88.9 | |
| CDistNet w/o TPSTraining Data=90K+ST2021.11 | 88.8 | |
| ABINet + InfoBatchPruned %=26.6%2025.03 | 88.8 | |
| CDistNetTraining Data=90K+ST2021.11 | 88.7 | |
| ABINetPruned %=0%2025.03 | 88.7 | |
| ABINet + SeTaPruned %=71.0%2025.03 | 88.2 | |
| ABINet + InfoBatchPruned %=50.3%2025.03 | 88.1 | |
| ABINet + InfoBatchPruned %=38.1%2025.03 | 88 | |
| ABINet + SeTaPruned %=40.4%2025.03 | 88 | |
| GTRTraining Data=90K+ST, Params=42.1M, Time=18.8ms2023.05 | 87.9 | |
| S-GTRYear=2022, Training Data=90K+ST2021.11 | 87.9 | |
| PREN2DYear=20212021.11 | 87.6 | |
| Pren2DTraining Data=90K+ST+Real, Time=67.4ms2023.05 | 87.6 | |
| Pren2DYear=2021, Training Data=90K+ST+Real2021.11 | 87.6 | |
| NRTRTraining Data=90K+ST, Params=31.7M, Time=212ms2023.05 | 86.6 | |
| NRTRYear=2019, Training Data=90K+ST2021.11 | 86.6 | |
| SATRNFeature map=2D, Training data=MJ+ST, Dictionary matching=False2019.10 | 86.5 | |
| SATRNYear=20202021.11 | 86.5 | |
| SATRNTraining Data=MJ and ST, Lexicon=None2022.09 | 86.5 | |
| SARLexicon-free=true, Character-level annotations=false2021.06 | 86.4 | |
| PETRTraining data=MJ+ST2023.12 | 86.2 | |
| VisionLANYear=20212021.11 | 86 | |
| VisionLANTraining Data=MJ and ST, Lexicon=None2022.09 | 86 | |
| VisionLANTraining Data=90K+ST, Params=32.8M, Time=28.0ms2023.05 | 86 | |
| VisionLANYear=2021, Training Data=90K+ST2021.11 | 86 | |
| VisionLANTraining data=MJ+ST2023.12 | 86 | |
| JVSRYear=20212021.11 | 85.7 | |
| JVSRYear=2021, Training Data=90K+ST2021.11 | 85.7 | |
| SRNBackbone=ResNet, Training Data=Synth90K+SynthText, Annotations=word-level, Lexicon=None2020.03 | 85.1 | |
| SRNYear=20202021.11 | 85.1 | |
| SRNTraining Data=MJ and ST, Lexicon=None2022.09 | 85.1 | |
| SRNTraining Data=90K+ST, Params=49.3M, Time=26.9ms2023.05 | 85.1 | |
| SRNYear=2020, Training Data=90K+ST2021.11 | 85.1 | |
| SRNTraining data=MJ+ST2023.12 | 85.1 | |
| TextScannerTraining data=MJ+ST2023.12 | 84.8 | |
| MASTERLexicon-free=true, Character-level annotations=false2021.06 | 84.5 | |
| SCATTERLexicon-free=true, Character-level annotations=false2021.06 | 84.5 | |
| MASTERTraining Data=MJ and ST, Lexicon=None2022.09 | 84.5 | |
| TextScannerTraining Data=MJ and ST, Lexicon=None2022.09 | 84.3 | |
| TextscannerYear=2020, Training Data=90K+ST2021.11 | 84.3 | |
| TPS-ResNet-BiLSTM-AttnTrain Dataset=S90k+ST+TextOCR2021.05 | 84.19 | |
| STAR-NetTrain Dataset=S90k+ST+TextOCR2021.05 | 83.88 | |
| SAR 3 x 20Lexicon-free=true, Character-level annotations=false2021.06 | 83.7 | |
| Mask TextSpotterLexicon-free=true, Character-level annotations=true2021.06 | 83.6 | |
| RCEEDLexicon-free=true, Character-level annotations=false2021.06 | 83.6 | |
| GA-SPINYear=20212021.11 | 83.2 | |
| SPINTraining Data=90K+ST2023.05 | 83.2 | |
| SGBANetTraining Data=90K+ST2023.05 | 83.1 | |
| RobustScannerLexicon-free=true, Character-level annotations=false2021.06 | 82.9 | |
| Lyu et al. (Parallel)Backbone=ResNet, Training Data=Synth90K+SynthText, Annotations=word-level, Lexicon=None2020.03 | 82.3 | |
| 2D-AttentionYear=2019, Training Data=90K+ST2021.11 | 82.3 | |
| Liao et al. (SAM)Backbone=ResNet, Training Data=Synth90K+SynthText, Annotations=word-level, Lexicon=None2020.03 | 82.2 | |
| STANTraining Data=90K+ST2023.05 | 82.2 | |
| MaskTextSpotterYear=2019, Training Data=90K+ST2021.11 | 82.2 | |
| ViTSTRTraining Data=MJ and ST, Lexicon=None2022.09 | 81.8 | |
| ViTSTR-BTraining data=MJ+ST2023.12 | 81.8 | |
| AutoSTRYear=20202021.11 | 81.7 |