Scene Text Recognition on IC15 (test)
93.5Word AccuracyDTrOCR
Evaluation Results
| Method | Links | |
|---|---|---|
| DTrOCRTraining data=Real2023.08 | 93.5 | |
| DTrOCRTraining data=Real2023.08 | 93.2 | |
| CLIP4STR-L*Training data=REBU-Syn2023.12 | 92.6 | |
| PARSeq-BTrain Data=REBU-Syn2023.12 | 92.44 | |
| PARSeq-LTrain Data=REBU-Syn2023.12 | 92.32 | |
| DTrOCRTraining data=Synth2023.08 | 92.3 | |
| CLIP4STR-B*Training data=REBU-Syn2023.12 | 92.27 | |
| PARSeq-SParameters=22M, Train Data=REBU-Syn2023.12 | 92.27 | |
| CLIP4STR-LTraining data=Real2023.12 | 91.66 | |
| PARSeq-HParameters=0.6B, Train Data=REBU-Syn2023.12 | 91.66 | |
| CLIP4STR-BTraining data=Real2023.12 | 91.39 | |
| PARSeq_ATrain data=R2022.07 | 90.7 | |
| PARSeqTraining data=Real2023.08 | 90.7 | |
| MAERec-BTraining data=Union14M-L2023.12 | 90.7 | |
| ABINetTraining data=Real2023.12 | 90.5 | |
| DTrOCRTraining data=Synth2023.08 | 90.4 | |
| PARSeqATraining data=Real, Reproduced=true2023.12 | 90.34 | |
| ABINetTrain data=R2022.07 | 90.2 | |
| ABINetTraining data=Real2023.08 | 90.2 | |
| MaskOCR(ViT-L)Training data=MJ+ST2023.12 | 90.1 | |
| TRBATrain data=R2022.07 | 89.8 | |
| TRBATraining data=Real2023.08 | 89.8 | |
| PARSeq_NTrain data=R2022.07 | 89.6 | |
| PARSeq_ATrain data=R2022.07 | 89.6 | |
| PARSeqTraining data=Real2023.08 | 89.6 | |
| ViTSTR-STraining data=Real2023.12 | 89 | |
| DiG-ViT-BTraining data=Real2023.12 | 88.9 | |
| TRBATrain data=R2022.07 | 88.7 | |
| MaskOCRBackbone=ViT-L2021.09 | 88.7 | |
| MaskOCR_LARGETraining data=Synth2023.08 | 88.7 | |
| TRBATraining data=Real2023.08 | 88.7 | |
| ABINetTrain data=R2022.07 | 88.5 | |
| ABINetTraining data=Real2023.08 | 88.5 | |
| VITSTR-STrain data=R2022.07 | 88.4 | |
| PARSeq_NTrain data=R2022.07 | 88.4 | |
| TrOCRBackbone=TrOCR-Large, Training Data=Synthetic + Benchmark2021.09 | 88.1 | |
| MaskOCRBackbone=ViT-B2021.09 | 87.3 | |
| MaskOCR_BASETraining data=Synth2023.08 | 87.3 | |
| MaskOCR(ViT-B)Training data=MJ+ST2023.12 | 87.3 | |
| VITSTR-STrain data=R2022.07 | 87.1 | |
| DiG-ViT-BTraining data=MJ+ST2023.12 | 87.1 | |
| TrOCRBackbone=TrOCR-Base, Training Data=Synthetic + Benchmark2021.09 | 86.9 | |
| SVTR_LARGETraining data=Synth2023.08 | 86.6 | |
| PARSeq_ATrain data=S2022.07 | 86.5 | |
| PARSeq2021.09 | 86.5 | |
| PARSeqTraining Data=90K+ST, Params=23.8M, Time=11.8ms2023.05 | 86.5 | |
| ABINet-LV+TPS++Training Data=90K+ST, Params=37.2M, Time=41.5ms2023.05 | 86.5 | |
| PARSeqTraining data=Synth2023.08 | 86.5 | |
| STN-CSTRTrain data=S2022.07 | 86.1 | |
| STN-CSTR2021.09 | 86.1 | |
| STAR-NetTrain Dataset=S90k+ST+TextOCR2021.05 | 86.09 | |
| ABINetTrain data=S32022.07 | 86 | |
| ABINet2021.09 | 86 | |
| ABINet-LVTraining Data=90K+ST, Params=36.7M, Time=37.2ms2023.05 | 86 | |
| ABINetTraining data=Synth2023.08 | 86 | |
| DiffusionSTRTraining data=Synth2023.08 | 86 | |
| ABINetYear=2021, Training Data=90K+ST2021.11 | 86 | |
| CDistNetTraining Data=90K+ST2021.11 | 86 | |
| ABINetTraining data=MJ+ST2023.12 | 86 | |
| NRTR+TPS++Training Data=90K+ST, Params=35.5M, Time=218ms2023.05 | 85.7 | |
| ABINet + SeTaPruned %=28.1%2025.03 | 85.6 | |
| ABINetPruned %=0%2025.03 | 85.4 | |
| ABINet + InfoBatchPruned %=26.6%2025.03 | 85.4 | |
| ABINet + SeTaPruned %=40.4%2025.03 | 85.4 | |
| SVTR_BASETraining data=Synth2023.08 | 85.2 | |
| CDistNet w/o TPSTraining Data=90K+ST2021.11 | 85.2 | |
| TPS-ResNet-BiLSTM-AttnTrain Dataset=S90k+ST+TextOCR2021.05 | 85.15 | |
| PARSeq_NTrain data=S2022.07 | 85.1 | |
| ABINet + SeTaPruned %=71.0%2025.03 | 84.9 | |
| CVAE-FeedTrain data=S2022.07 | 84.6 | |
| CVAE-Feed.2021.09 | 84.6 | |
| GTRTraining Data=90K+ST, Params=42.1M, Time=18.8ms2023.05 | 84.6 | |
| CVAE-FeedTraining data=Synth2023.08 | 84.6 | |
| S-GTRYear=2022, Training Data=90K+ST2021.11 | 84.6 | |
| TRBATrain data=S2022.07 | 84.3 | |
| ABINet + InfoBatchPruned %=38.1%2025.03 | 84.2 | |
| TrOCRBackbone=TrOCR-Large, Training Data=Synthetic + Benchmark2021.09 | 84.1 | |
| TrOCRLargeTraining data=MJ+ST+B2023.12 | 84.1 | |
| Bhunia et al.Train data=S2022.07 | 84 | |
| TrOCRBackbone=TrOCR-Large, Training Data=Synthetic2021.09 | 84 | |
| Bhunia2021.09 | 84 | |
| TrOCR_LARGETraining data=Synth2023.08 | 84 | |
| JVSRTraining data=Synth2023.08 | 84 | |
| JVSRYear=2021, Training Data=90K+ST2021.11 | 84 | |
| ABINet + InfoBatchPruned %=50.3%2025.03 | 84 | |
| VisionLANTrain data=S2022.07 | 83.7 | |
| VisionLAN2021.09 | 83.7 | |
| VisionLANTraining Data=90K+ST, Params=32.8M, Time=28.0ms2023.05 | 83.7 | |
| VisionLANTraining data=Synth2023.08 | 83.7 | |
| VisionLANYear=2021, Training Data=90K+ST2021.11 | 83.7 | |
| VisionLANTraining data=MJ+ST2023.12 | 83.7 | |
| TextScannerTrain data=S*2022.07 | 83.5 | |
| TextScanner2021.09 | 83.5 | |
| TextScannerTraining data=Synth2023.08 | 83.5 | |
| TextScannerTraining data=MJ+ST2023.12 | 83.5 | |
| PETRTraining data=MJ+ST2023.12 | 83.3 | |
| ABINetTrain data=S2022.07 | 83.1 | |
| PREN2DTrain data=S2022.07 | 83 | |
| PREN2D2021.09 | 83 | |
| Pren2DTraining Data=90K+ST+Real, Time=67.4ms2023.05 | 83 |