Scene Text Recognition on 6 common benchmarks (test)
99.3Word Accuracy (IIIT)MNSP
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| MNSPlabel_correction=true2026.05 | 99.3 | — | 98.5 | 97.2 | 91.4 | 95.2 | 98.3 | 96.7 | |
| SVTRv2-ARType=AR2026.06 | 99.23 | — | 98.72 | 97.06 | 90.61 | 95.35 | 97.92 | 96.48 | |
| SVTRv2Type=CTC2026.06 | 99.17 | — | 99.18 | 98.45 | 90.23 | 94.73 | 98.96 | 96.79 | |
| WATERecType=AR2026.06 | 99.17 | — | 98.72 | 97.68 | 91.11 | 95.19 | 98.26 | 96.69 | |
| MAERecType=AR2026.06 | 99.13 | — | 99.07 | 97.99 | 91.05 | 96.28 | 98.26 | 96.96 | |
| SSM-S2026.05 | 99 | — | 98.3 | 97.8 | 89.5 | 94 | 98.3 | 96.1 | |
| CLIP-OCRDatasets=RL, P(M)=31.12024.11 | 98.8 | — | 98.2 | 98.3 | 88.8 | 94.3 | 97.6 | 95.5 | |
| NSPlabel_correction=true2026.05 | 98.8 | — | 98.6 | 96.9 | 90.8 | 93.3 | 97.6 | 96 | |
| ABINetType=PD2026.06 | 98.73 | — | 97.67 | 97.37 | 90.01 | 94.73 | 98.26 | 96.13 | |
| LPVDatasets=RL, P(M)=35.12024.11 | 98.7 | — | 97.9 | 98.6 | 88.7 | 93.9 | 96.2 | 95.4 | |
| DiG-S2026.05 | 98.7 | — | 97.8 | 98.5 | 88.9 | 92.7 | 96.5 | 95.5 | |
| MNSP2026.05 | 98.7 | — | 98.4 | 96.6 | 90.5 | 94.3 | 97.9 | 96.1 | |
| BaselineDatasets=RL, P(M)=24.42024.11 | 98.6 | — | 98.3 | 98 | 89.4 | 94.3 | 96.5 | 95.7 | |
| ParSeqDatasets=RL, P(M)=23.82024.11 | 98.5 | — | 98 | 97.7 | 88.7 | 94.7 | 96.5 | 95.3 | |
| TRBA-crDatasets=RL+RU, P(M)=49.92024.11 | 98.5 | — | 97.7 | 98.3 | 89.8 | 94.3 | 96.5 | 95.6 | |
| MAERec-BDatasets=RL+RU, P(M)=135.52024.11 | 98.5 | — | 98.1 | 97.8 | 89.5 | 94.4 | 98.6 | 95.6 | |
| ViSuDatasets=RL+RU, P(M)=24.42024.11 | 98.5 | — | 97.8 | 98.3 | 90.4 | 96.3 | 97.6 | 96 | |
| LMIM2026.05 | 98.5 | — | 98 | 97.7 | 88.7 | 94 | 98.6 | 95.9 | |
| PARSeqType=AR2026.06 | 98.4 | — | 98.13 | 97.68 | 89.4 | 95.5 | 97.57 | 96.11 | |
| MATRNDatasets=RL, P(M)=44.22024.11 | 98.2 | — | 97.9 | 96.9 | 88.2 | 94.1 | 97.9 | 95 | |
| MATRN2026.05 | 98.2 | — | 97.9 | 96.9 | 88.2 | 94.1 | 97.9 | 95.5 | |
| NSP2026.05 | 98.2 | — | 98.5 | 96.3 | 89.3 | 92.7 | 97.2 | 95.4 | |
| LISTERDatasets=RL, P(M)=49.92024.11 | 98.1 | — | 97.2 | 97.7 | 86.3 | 91.8 | 94.8 | 94.1 | |
| MAERec-SDatasets=RL+RU, P(M)=34.02024.11 | 98 | — | 97.6 | 96.8 | 87.1 | 93.2 | 97.9 | 94.5 | |
| PARSeq2026.05 | 98 | — | 96.8 | 95.2 | 85.2 | 90.5 | 96.5 | 93.8 | |
| MaskOCR-S2026.05 | 98 | — | 97.8 | 96.9 | 90.2 | 94.9 | 96.2 | 95.6 | |
| CCDrecalculated=true2026.05 | 98 | — | 98.3 | 96.4 | 90.3 | 92.7 | 98.3 | 95.6 | |
| MAERec-S2026.05 | 98 | — | 97.6 | 96.8 | 87.1 | 93.2 | 97.9 | 95.1 | |
| BUSNetType=PD2026.06 | 98 | — | 97.55 | 95.52 | 88.79 | 93.33 | 95.83 | 94.84 | |
| LPVType=PD2026.06 | 97.93 | — | 96.62 | 95.05 | 87.69 | 91.47 | 96.18 | 94.16 | |
| OTEType=AR2026.06 | 97.87 | — | 97.2 | 96.75 | 89.18 | 93.33 | 96.53 | 95.14 | |
| GIT2Fine-tuning Data=MJ+ST2022.05 | 97.6 | — | 97.8 | 95.5 | 85.6 | 91.3 | 99 | 94.5 | |
| ViSuDatasets=SL+RU, P(M)=24.42024.11 | 97.6 | — | 98.3 | 96.1 | 89.3 | 91.3 | 92.4 | 94.7 | |
| CLIP-OCR* [43]Datasets=SL, P(M)=31.12024.11 | 97.3 | — | 97.7 | 94.7 | 87.2 | 89.9 | 93.1 | 93.8 | |
| LPV* [52]Datasets=SL, P(M)=35.12024.11 | 97.3 | — | 97.6 | 94.6 | 87.5 | 90.9 | 94.8 | 94 | |
| SVTRType=CTC2026.06 | 97.3 | — | 96.38 | 94.74 | 87.47 | 89.3 | 93.75 | 93.16 | |
| ABINetDatasets=RL, P(M)=36.72024.11 | 97.2 | — | 97.2 | 95.7 | 87.6 | 92.1 | 94.4 | 93.9 | |
| MGPDatasets=RL, P(M)=141.12024.11 | 97.2 | — | 96.8 | 97.7 | 87.2 | 93.6 | 94.8 | 94.1 | |
| ABINet2026.05 | 97.2 | — | 97.2 | 95.7 | 87.6 | 92.1 | 94.4 | 94 | |
| ParSeq† [2]Datasets=SL, P(M)=23.82024.11 | 97 | — | 97 | 93.6 | 86.5 | 88.9 | 92.3 | 93.3 | |
| CPPDType=PD2026.06 | 97 | — | 97.43 | 96.29 | 87.02 | 92.71 | 96.18 | 94.44 | |
| LISTER* [5]Datasets=SL, P(M)=49.92024.11 | 96.9 | — | 97.9 | 93.8 | 87.5 | 89.6 | 90.6 | 93.5 | |
| ABINet* [10]Datasets=SL+Wiki+Uber, P(M)=36.72024.11 | 96.9 | — | 97 | 94.7 | 85.9 | 89 | 89.9 | 93 | |
| RobustScannerDatasets=RL2024.11 | 96.8 | — | 95.7 | 92.4 | 86.4 | 83.9 | 93.8 | 92.3 | |
| MATRN [25]Datasets=SL, P(M)=44.22024.11 | 96.6 | — | 97.9 | 95 | 86.6 | 90.6 | 93.5 | 93.5 | |
| BaseDescription=Baseline on original images2025.03 | 96.5 | — | 95.1 | 94.7 | 85.9 | 89.5 | 89.6 | 91.8 | |
| MaskOCRFine-tuning Data=MJ+ST2022.05 | 96.5 | — | 97.8 | 94.1 | 88.7 | 90.2 | 92.7 | 93.8 | |
| TRBA-cr* [54]Datasets=SL+RU2, P(M)=49.92024.11 | 96.5 | — | 98.3 | 96.3 | 89.3 | 93.3 | 93.4 | 94.5 | |
| MGP* [41]Datasets=SL, P(M)=141.12024.11 | 96.4 | — | 97.3 | 94.7 | 87.2 | 91 | 90.3 | 93.3 | |
| TRBA-ViSuDatasets=SL+RU, P(M)=49.92024.11 | 96.4 | — | 97.2 | 96.1 | 86.2 | 87.8 | 89.9 | 92.9 | |
| VisionLANDatasets=RL, P(M)=32.82024.11 | 96.3 | — | 95.1 | 91.3 | 83.6 | 85.4 | 92.4 | 91.2 | |
| VisionLAN2026.05 | 96.3 | — | 95.1 | 91.3 | 83.6 | 85.4 | 92.4 | 91.3 | |
| ABINetFine-tuning Data=MJ+ST2022.05 | 96.2 | — | 97.4 | 93.5 | 86 | 89.3 | 89.2 | 91.9 | |
| ABINet [10]Datasets=SL+Wiki, P(M)=36.72024.11 | 96.2 | — | 97.4 | 93.5 | 86 | 89.3 | 89.2 | 92.3 | |
| BaselineDatasets=SL, P(M)=24.42024.11 | 96.1 | — | 96.5 | 94.4 | 86 | 88.2 | 88.5 | 92.5 | |
| SVTR [9]Datasets=SL, P(M)=24.62024.11 | 96 | — | 97.1 | 91.5 | 85.2 | 89.9 | 91.7 | 92.3 | |
| SVTRDatasets=RL, P(M)=24.62024.11 | 95.9 | — | 95.5 | 92.4 | 83.9 | 85.7 | 93.1 | 91.3 | |
| S-GTRFine-tuning Data=MJ+ST2022.05 | 95.8 | — | 96.8 | 94.1 | 84.6 | 87.9 | 92.3 | 91.9 | |
| VisionLAN [42]Datasets=SL, P(M)=32.82024.11 | 95.8 | — | 95.7 | 91.7 | 83.7 | 86 | 88.5 | 91.2 | |
| TRBA†Datasets=SL, P(M)=49.92024.11 | 95.6 | — | 96.3 | 92.7 | 83.4 | 85.6 | 87.9 | 91.2 | |
| TRBA-cr† [54]Datasets=SL+RU, P(M)=49.92024.11 | 95.5 | — | 97.4 | 94.6 | 87.1 | 90.1 | 91.3 | 92.9 | |
| SRNDatasets=RL, P(M)=54.72024.11 | 95.5 | — | 94.7 | 89.5 | 79.1 | 83.9 | 91.3 | 89.3 | |
| Ro.ScannerFine-tuning Data=MJ+ST2022.05 | 95.3 | — | 94.8 | 88.1 | 77.1 | 79.5 | 90.3 | 87.5 | |
| GITFine-tuning Data=MJ+ST2022.05 | 95.3 | — | 97.3 | 95.2 | 83.7 | 89.9 | 96.2 | 92.9 | |
| RobustScanner [50]Datasets=SL2024.11 | 95.3 | — | 94.8 | 88.1 | 77.1 | 79.5 | 90.3 | 88.4 | |
| TRBA-pr* [1]Datasets=RL+RU1, P(M)=49.92024.11 | 94.9 | — | 94.2 | 92.4 | 84.8 | 82.8 | 87.9 | 90.7 | |
| TRBA-prDatasets=RL1+RU1, P(M)=49.92024.11 | 94.9 | — | 94.2 | 92.4 | 84.8 | 82.8 | 87.9 | 90.7 | |
| SRNFine-tuning Data=MJ+ST2022.05 | 94.8 | — | 95.5 | 91.5 | 82.7 | 85.1 | 87.8 | 89.6 | |
| SRN [49]Datasets=SL, P(M)=54.72024.11 | 94.8 | — | 95.5 | 91.5 | 82.7 | 85.1 | 87.8 | 90.4 | |
| SAMFine-tuning Data=MJ+ST2022.05 | 93.9 | — | 95.3 | 90.6 | 77.3 | 82.2 | 87.8 | 87.8 | |
| CRNNType=CTC2026.06 | 93.07 | — | 92.3 | 87.02 | 80.95 | 80.47 | 82.64 | 86.07 | |
| GITFine-tuning Data=TextCaps2022.05 | 92.9 | — | 94.2 | 91.5 | 78.2 | 87.1 | 95.5 | 89.9 | |
| CRNNDatasets=RL, P(M)=8.32024.11 | 90.8 | — | 91.8 | 83.8 | 71.8 | 70.4 | 80.9 | 83.1 | |
| CRNN-pr* [1]Datasets=RL+RU1, P(M)=8.52024.11 | 90.2 | — | 91.5 | 86.1 | 77.8 | 74.1 | 81.6 | 85.1 | |
| CRNN-prDatasets=RL1+RU1, P(M)=8.52024.11 | 90.2 | — | 91.5 | 86.1 | 77.8 | 74.1 | 81.6 | 85.1 | |
| CRNN-ViSuDatasets=SL+RU, P(M)=8.62024.11 | 90 | — | 90.9 | 84.7 | 71.1 | 73.6 | 77.4 | 83 | |
| CRNN†Datasets=SL, P(M)=8.62024.11 | 88.8 | — | 90.3 | 82.7 | 69.5 | 69.3 | 76.7 | 81.4 | |
| CRNN [33]Datasets=SL, P(M)=8.32024.11 | 82.9 | — | 91.9 | 81.6 | 69.4 | 70 | 65.5 | 78.6 | |
| OursReal Training Dataset=Union14M-L2025.03 | 78.4 | — | 76.2 | 91.3 | 85.8 | 88.2 | 77.4 | 82.9 | |
| OursReal Training Dataset=MLT20172025.03 | 76.6 | — | 74.8 | 89.6 | 83 | 86.4 | 80.6 | 81.8 | |
| TextCtrlReal Training Dataset=MLT20172025.03 | 70 | — | 68.8 | 73.7 | 62.6 | 63.4 | 58.7 | 66.2 | |
| Ours (w/o Lcyc)Ablation=Without cycle-consistency loss2025.03 | 61.2 | — | 61.2 | 66.8 | 53.8 | 46.4 | 44.8 | 55.7 | |
| MOSTELReal Training Dataset=MLT20172025.03 | 48.2 | — | 40.6 | 41.1 | 34.4 | 21.6 | 35.1 | 36.8 | |
| GIT2022.05 | — | 92.9 | — | — | — | — | — | — | |
| GIT22022.05 | — | 94.5 | — | — | — | — | — | — | |
| PARSeq-SReal DataSet=R+E+B+U, Syn DataSet=Syn, Data Ratio=1:0.52023.12 | — | 96.19 | — | — | — | — | — | — | |
| PARSeq-SReal DataSet=R+E+B+U, Syn DataSet=MJ+ST, Data Ratio=1:2.52023.12 | — | 96.24 | — | — | — | — | — | — | |
| PARSeq-SReal DataSet=R+E+B+U, Syn DataSet=MJ+ST+Syn, Data Ratio=1:32023.12 | — | 96.85 | — | — | — | — | — | — | |
| Prior SOTA2022.05 | — | 93.8 | — | — | — | — | — | — |