Text Detection on Total-Text (test)
90F-MeasureESTextSpotter
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ESTextSpotterParadigm=Supervised, Data Setting=Fully Labeled2025.04 | 90 | — | — | — | |
| UNITSModel Type=Fine-tuned2023.04 | 89.8 | — | — | — | |
| TPSNetexternal dataset pretrain=true, end-to-end spotting training=true2021.10 | 88.5 | 90.2 | 86.8 | 14.3 | |
| UNITS SharedModel Type=Unified2023.04 | 88.4 | — | — | — | |
| TPSNetexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 88.1 | 89.5 | 86.8 | 14.3 | |
| GLASSParadigm=Supervised, Data Setting=Fully Labeled2025.04 | 88.1 | — | — | — | |
| SwinTextSpotterBackbone=Dilated Swin-Transformer2022.03 | 88 | — | — | — | |
| SwinTextSpotter2023.04 | 88 | — | — | — | |
| SwinTextSpotterParadigm=Supervised, Data Setting=Fully Labeled2025.04 | 88 | — | — | — | |
| Unified DetectorMode=word, Training Data (Pub)=true, Training Data (HierText)=false2022.03 | 87.94 | 85.49 | 90.53 | — | |
| Unified DetectorMode=word, Training Data (Pub)=true, Training Data (HierText)=true2022.03 | 87.9 | 84.96 | 91.06 | — | |
| BPNexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 87.9 | 90.7 | 85.2 | 10.7 | |
| ABPNetVenue=ICCV21, Training Data (Pub)=false, Training Data (HierText)=false2022.03 | 87.85 | 90.67 | 85.19 | — | |
| FAST-B-800#Param (total)=10.6M, Scale=S: 800, FLOPS=100.1G, PyTorch FPS (1080Ti)=46, PyTorch FPS (V100)=69.2, TensorRT FPS (V100 FP16) BS=1=143, TensorRT FPS (V100 FP16) BS=2=177.5, TensorRT FPS (V100 FP16) BS=4=189.2, TensorRT FPS (V100 FP16) BS=8=195.32021.11 | 87.5 | — | — | — | |
| Unbiased Teacher v2*Paradigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 87.5 | — | — | — | |
| SemiETSParadigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 87.5 | — | — | — | |
| CRAFTS2023.04 | 87.4 | — | — | — | |
| DeepSoloParadigm=Supervised, Data Setting=Fully Labeled2025.04 | 87.3 | — | — | — | |
| Supervised baselineParadigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 87.3 | — | — | — | |
| SwinTextSpotterBackbone=ResNet-502022.03 | 87.2 | — | — | — | |
| ABCNet v22022.03 | 87 | — | — | — | |
| Boundaryexternal dataset pretrain=true, end-to-end spotting training=true2021.10 | 87 | 88.9 | 85 | — | |
| ABCNetV2external dataset pretrain=true, end-to-end spotting training=true2021.10 | 87 | 90.2 | 84.1 | — | |
| ABCNet v22023.04 | 87 | — | — | — | |
| ABCNet v2Paradigm=Supervised, Data Setting=Fully Labeled2025.04 | 87 | — | — | — | |
| TESTR2023.04 | 86.9 | — | — | — | |
| TESTRParadigm=Supervised, Data Setting=Fully Labeled2025.04 | 86.9 | — | — | — | |
| Mean-Teacher*Paradigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 86.9 | — | — | — | |
| TPSNetexternal dataset pretrain=false, end-to-end spotting training=false2021.10 | 86.6 | 89.2 | 84 | 14.3 | |
| FAST-B-640#Param (total)=10.6M, Scale=S: 640, FLOPS=64.0G, PyTorch FPS (1080Ti)=67.5, PyTorch FPS (V100)=99.8, TensorRT FPS (V100 FP16) BS=1=219.7, TensorRT FPS (V100 FP16) BS=2=261.8, TensorRT FPS (V100 FP16) BS=4=293.3, TensorRT FPS (V100 FP16) BS=8=2972021.11 | 86.4 | — | — | — | |
| CT-640external training data=true, resolution=6402021.07 | 86.3 | 90.5 | 82.5 | 40 | |
| CentripetalTextVenue=NeurIPS21, Training Data (Pub)=false, Training Data (HierText)=false2022.03 | 86.3 | 90.6 | 82.5 | — | |
| PGNet2022.03 | 86.1 | — | — | — | |
| Soft-Teacher*Paradigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 86.1 | — | — | — | |
| FCENetexternal dataset pretrain=false, end-to-end spotting training=false2021.10 | 85.8 | 89.3 | 82.5 | — | |
| FAST-B-512#Param (total)=10.6M, Scale=S: 512, FLOPS=41.0G, PyTorch FPS (1080Ti)=93.2, PyTorch FPS (V100)=128.6, TensorRT FPS (V100 FP16) BS=1=281, TensorRT FPS (V100 FP16) BS=2=358.6, TensorRT FPS (V100 FP16) BS=4=414.3, TensorRT FPS (V100 FP16) BS=8=4482021.11 | 85.8 | — | — | — | |
| DRRGexternal training data=true, venue=CVPR'202021.07 | 85.7 | 86.5 | 84.9 | — | |
| DRRGNexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 85.7 | 86.5 | 84.9 | — | |
| DEER2023.04 | 85.7 | — | — | — | |
| CT-512external training data=true, resolution=5122021.07 | 85.6 | 90.2 | 81.5 | 57 | |
| CharNet2022.03 | 85.6 | — | — | — | |
| ContourNetexternal dataset pretrain=false, end-to-end spotting training=false2021.10 | 85.4 | 86.9 | 83.9 | 3.8 | |
| FAST-S-640#Param (total)=9.7M, Scale=S: 640, FLOPS=51.7G, PyTorch FPS (1080Ti)=85.3, PyTorch FPS (V100)=115.9, TensorRT FPS (V100 FP16) BS=1=249.6, TensorRT FPS (V100 FP16) BS=2=301.7, TensorRT FPS (V100 FP16) BS=4=325, TensorRT FPS (V100 FP16) BS=8=3322021.11 | 85.4 | — | — | — | |
| Text PerceptronStage=End-to-end2020.02 | 85.2 | 88.8 | 81.8 | — | |
| Mask TextSpotter2022.03 | 85.2 | — | — | — | |
| PCRVenue=CVPR21, Training Data (Pub)=true, Training Data (HierText)=false2022.03 | 85.2 | 88.5 | 82 | — | |
| MaskTextSpotterexternal dataset pretrain=true, end-to-end spotting training=true2021.10 | 85.2 | 88.3 | 82.4 | — | |
| PCRexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 85.2 | 88.5 | 82 | — | |
| PAN-640Ext.=true, Short side scale=6402019.08 | 85 | 89.3 | 81 | 39.6 | |
| PANexternal training data=true, venue=ICCV'192021.07 | 85 | 89.3 | 81 | 39.6 | |
| PANexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 85 | 89.3 | 81 | 39.6 | |
| PAN-640Scale=S: 640, FLOPS=65.4G, PyTorch FPS (1080Ti)=39.6, PyTorch FPS (V100)=64.6, TensorRT FPS (V100 FP16) BS=1=130.4, TensorRT FPS (V100 FP16) BS=2=154.4, TensorRT FPS (V100 FP16) BS=4=162.8, TensorRT FPS (V100 FP16) BS=8=180.72021.11 | 85 | — | — | — | |
| CT-640external training data=false, resolution=6402021.07 | 84.9 | 88.8 | 81.4 | 40 | |
| FAST-S-512#Param (total)=9.7M, Scale=S: 512, FLOPS=33.1G, PyTorch FPS (1080Ti)=115.5, PyTorch FPS (V100)=150.5, TensorRT FPS (V100 FP16) BS=1=289.7, TensorRT FPS (V100 FP16) BS=2=400.2, TensorRT FPS (V100 FP16) BS=4=465.7, TensorRT FPS (V100 FP16) BS=8=481.52021.11 | 84.9 | — | — | — | |
| DBexternal training data=true, venue=AAAI'202021.07 | 84.7 | 87.1 | 82.5 | 32 | |
| DBexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 84.7 | 87.1 | 82.5 | 32 | |
| PAN-512Ext.=true, Short side scale=5122019.08 | 84.3 | 89.4 | 79.7 | 57.1 | |
| Boundaryexternal training data=true, venue=AAAI'202021.07 | 84.3 | 85.2 | 83.5 | — | |
| PAN-512#Param (total)=12.2M, Scale=S: 512, FLOPS=41.9G, PyTorch FPS (1080Ti)=57.1, PyTorch FPS (V100)=85.5, TensorRT FPS (V100 FP16) BS=1=187.6, TensorRT FPS (V100 FP16) BS=2=204.3, TensorRT FPS (V100 FP16) BS=4=215.5, TensorRT FPS (V100 FP16) BS=8=243.82021.11 | 84.3 | — | — | — | |
| CT-512external training data=false, resolution=5122021.07 | 84.2 | 87.9 | 80.8 | 57 | |
| FAST-T-640#Param (total)=8.5M, Scale=S: 640, FLOPS=46.0G, PyTorch FPS (1080Ti)=95.5, PyTorch FPS (V100)=136.4, TensorRT FPS (V100 FP16) BS=1=275.6, TensorRT FPS (V100 FP16) BS=2=287.1, TensorRT FPS (V100 FP16) BS=4=345.4, TensorRT FPS (V100 FP16) BS=8=3642021.11 | 84.2 | — | — | — | |
| Semi-DETR*Paradigm=Semi-supervised, Data Setting=Fully Labeled2025.04 | 84 | — | — | — | |
| CRAFTexternal training data=true, venue=CVPR'192021.07 | 83.6 | 87.6 | 79.9 | — | |
| CRAFTVenue=CVPR19, Training Data (Pub)=false, Training Data (HierText)=false2022.03 | 83.6 | 87.6 | 79.9 | — | |
| CRAFTexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 83.6 | 87.6 | 79.9 | — | |
| PAN-640Ext.=false, Short side scale=6402019.08 | 83.5 | 88 | 79.4 | 39.6 | |
| PANexternal training data=false, venue=ICCV'192021.07 | 83.5 | 88 | 79.4 | 39.6 | |
| FAST-T-512#Param (total)=8.5M, Scale=S: 512, FLOPS=29.5G, PyTorch FPS (1080Ti)=131.1, PyTorch FPS (V100)=176.4, TensorRT FPS (V100 FP16) BS=1=343.4, TensorRT FPS (V100 FP16) BS=2=408.2, TensorRT FPS (V100 FP16) BS=4=498.5, TensorRT FPS (V100 FP16) BS=8=525.62021.11 | 83.5 | — | — | — | |
| Text PerceptronStage=2-stage2020.02 | 83.3 | 88.1 | 78.9 | — | |
| LOMOexternal training data=true, venue=CVPR'19, multi-scale testing=true2021.07 | 83.3 | 87.6 | 79.3 | — | |
| SPCNET2018.11 | 82.9 | 83 | 82.8 | — | |
| SPCNetExt.=true, Venue=AAAI'192019.08 | 82.9 | 82.8 | 82.9 | — | |
| SPCNet2020.02 | 82.9 | 83 | 82.8 | — | |
| SPCNetexternal training data=true, venue=AAAI'192021.07 | 82.9 | 83 | 82.8 | — | |
| DB-R18#Param (total)=13.8M, Scale=H: 640, FLOPS=30.0G, PyTorch FPS (1080Ti)=50, PyTorch FPS (V100)=73, TensorRT FPS (V100 FP16) BS=1=105.4, TensorRT FPS (V100 FP16) BS=2=143.7, TensorRT FPS (V100 FP16) BS=4=204.5, TensorRT FPS (V100 FP16) BS=8=227.62021.11 | 82.8 | — | — | — | |
| PAN-512Ext.=false, Short side scale=5122019.08 | 82.4 | 86.7 | 78.4 | 57.1 | |
| LOMO2020.02 | 81.6 | 75.7 | 88.6 | — | |
| LOMOexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 81.6 | 88.6 | 75.7 | 4.4 | |
| FAST-T-448#Param (total)=8.5M, Scale=S: 448, FLOPS=22.6G, PyTorch FPS (1080Ti)=152.8, PyTorch FPS (V100)=187, TensorRT FPS (V100 FP16) BS=1=385.9, TensorRT FPS (V100 FP16) BS=2=524.8, TensorRT FPS (V100 FP16) BS=4=628.1, TensorRT FPS (V100 FP16) BS=8=634.72021.11 | 81.6 | — | — | — | |
| SegLink++external training data=true, venue=PR'192021.07 | 81.5 | 82.1 | 80.9 | — | |
| FTSN2018.11 | 81.3 | 84.7 | 78 | — | |
| FTSN2020.02 | 81.3 | 84.7 | 78 | — | |
| CT-320external training data=true, resolution=3202021.07 | 81.2 | 88 | 75.4 | 93.2 | |
| Baseline2018.11 | 81 | 81.5 | 80.5 | — | |
| PSENet-1sExt.=true, Venue=CVPR'192019.08 | 80.9 | 84 | 78 | 3.9 | |
| PSENet-1s2020.02 | 80.9 | 84 | 78 | — | |
| PSENetexternal training data=true, venue=CVPR'192021.07 | 80.9 | 84 | 78 | 3.9 | |
| PSENetVenue=CVPR19, Training Data (Pub)=false, Training Data (HierText)=false2022.03 | 80.9 | 84 | 78 | — | |
| PSENetexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 80.9 | 84 | 78 | 3.9 | |
| TextField2020.02 | 80.6 | 81.2 | 79.9 | — | |
| TextRayexternal dataset pretrain=false, end-to-end spotting training=false2021.10 | 80.6 | 83.5 | 77.9 | — | |
| CSE2020.02 | 80.2 | 81.4 | 79.1 | — | |
| SASTexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 80.2 | 83.8 | 76.9 | — | |
| CSEexternal dataset pretrain=false, end-to-end spotting training=false2021.10 | 80.2 | 81.4 | 79.1 | 0.4 | |
| SAEexternal dataset pretrain=true, end-to-end spotting training=false2021.10 | 80.1 | 82.7 | 77.8 | — | |
| PAN-320Ext.=true, Short side scale=3202019.08 | 79.9 | 85.6 | 75 | 82.4 | |
| PAN-320Scale=S: 320, FLOPS=16.4G, PyTorch FPS (1080Ti)=82.4, PyTorch FPS (V100)=110.8, TensorRT FPS (V100 FP16) BS=1=237.5, TensorRT FPS (V100 FP16) BS=2=323.2, TensorRT FPS (V100 FP16) BS=4=363.4, TensorRT FPS (V100 FP16) BS=8=385.22021.11 | 79.9 | — | — | — | |
| PSENet-4s#Param (total)=28.7M, Scale=L: 1280, FLOPS=345.9G, PyTorch FPS (1080Ti)=8.4, PyTorch FPS (V100)=10.8, TensorRT FPS (V100 FP16) BS=1=59.3, TensorRT FPS (V100 FP16) BS=2=69.2, TensorRT FPS (V100 FP16) BS=4=72.8, TensorRT FPS (V100 FP16) BS=8=77.32021.11 | 79.6 | — | — | — | |
| CT-320external training data=false, resolution=3202021.07 | 79.4 | 87.6 | 72.7 | 93.2 | |
| MSRexternal training data=true, venue=IJCAI'192021.07 | 79 | 83.8 | 74.8 | — |