Table Recognition on PubTabNet (test)
98.28TEDS (Simple)TFLOP + GAP
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TFLOP + GAPloss_function=GAP loss2026.06 | 98.28 | — | 96.49 | — | — | — | |
| TFLOP2026.06 | 98.27 | — | 96.43 | — | — | — | |
| MuTabNet2024.04 | 98.01 | 94.98 | 96.53 | — | — | — | |
| VCGroup2021.07 | 97.9 | 94.68 | 96.32 | — | — | — | |
| VCGroup2024.04 | 97.9 | 94.68 | 96.32 | — | — | — | |
| Davar-Lab-OCR2021.07 | 97.88 | 94.78 | 96.36 | — | — | — | |
| Davar-Lab-OCR2024.04 | 97.88 | 94.78 | 96.36 | — | — | — | |
| Ly and Takasu [18]2024.04 | 97.77 | 94.58 | 96.21 | — | — | — | |
| TABLET2026.05 | 97.7 | — | — | — | — | — | |
| SPRINT2026.06 | 97.68 | — | — | — | — | — | |
| USTC-NELSLIP2021.07 | 97.6 | 94.89 | 96.27 | — | — | — | |
| XM2024.04 | 97.6 | 94.89 | 96.27 | — | — | — | |
| Ly and Takasu [19]2024.04 | 97.6 | 94.68 | 96.17 | — | — | — | |
| Ly et al.2024.04 | 97.51 | 94.37 | 95.97 | — | — | — | |
| TSRFormerTraining Dataset=PTN2023.03 | 97.5 | — | — | — | — | — | |
| TSRFormerFamily=Split-and-Merge2026.04 | 97.5 | — | — | — | — | — | |
| SEMv2Family=Split-and-Merge2026.04 | 97.5 | — | — | — | — | — | |
| UniTabNetFamily=Img-to-Seq2026.04 | 97.5 | — | — | — | — | — | |
| SEMv22026.05 | 97.5 | — | — | — | — | — | |
| TSRFormer2026.05 | 97.5 | — | — | — | — | — | |
| TableFormer2026.06 | 97.5 | — | 93.6 | — | — | — | |
| SLANET2026.06 | 97.48 | — | 95.89 | — | — | — | |
| DBJ2021.07 | 97.39 | 93.87 | 95.66 | — | — | — | |
| DBJ2024.04 | 97.39 | 93.87 | 95.66 | — | — | — | |
| YG2021.07 | 97.38 | 94.79 | 96.11 | — | — | — | |
| YG2024.04 | 97.38 | 94.79 | 96.11 | — | — | — | |
| PaodingAI2021.07 | 97.35 | 93.79 | 95.61 | — | — | — | |
| PaodingAI2024.04 | 97.35 | 93.79 | 95.61 | — | — | — | |
| TAL2021.07 | 97.3 | 93.93 | 95.65 | — | — | — | |
| TAL2024.04 | 97.3 | 93.93 | 95.65 | — | — | — | |
| VASTTraining Dataset=PTN2023.03 | 97.23 | — | 96.31 | — | — | — | |
| VASTFamily=Img-to-Seq2026.04 | 97.23 | — | 96.31 | — | — | — | |
| VAST2026.06 | 97.23 | — | 96.31 | — | — | — | |
| VAST2026.05 | 97.2 | — | — | — | — | 1.38 | |
| LTIAYN2021.07 | 97.18 | 92.4 | 94.84 | — | — | — | |
| LTIAYN2024.04 | 97.18 | 92.4 | 94.84 | — | — | — | |
| TRUSTTraining Dataset=PTN2023.03 | 97.1 | — | 96.2 | — | — | — | |
| TRUSTFamily=Split-and-Merge2026.04 | 97.1 | — | 96.2 | — | — | — | |
| TRUST2026.05 | 97.1 | — | — | — | — | 10 | |
| RobustTabNetTraining Dataset=PTN2023.03 | 97 | — | — | — | — | — | |
| GridFormer2026.05 | 97 | — | — | — | — | — | |
| anyone2021.07 | 96.95 | 93.43 | 95.23 | — | — | — | |
| anyone2024.04 | 96.95 | 93.43 | 95.23 | — | — | — | |
| TableSeqFamily=Img-to-Seq2026.04 | 96.83 | — | 95.23 | — | — | — | |
| FastTab2026.05 | 96.8 | — | — | — | — | 44.89 | |
| TableFormerTraining Dataset=PTN2023.03 | 96.75 | — | 93.6 | — | — | — | |
| TableFormerFamily=Img-to-Seq2026.04 | 96.75 | — | 93.6 | — | — | — | |
| LGPMATraining Dataset=PTN2023.03 | 96.7 | — | 94.6 | — | — | — | |
| LGPMAFamily=Bottom-Up2026.04 | 96.7 | — | 94.6 | — | — | — | |
| LGPMA2026.05 | 96.7 | — | — | — | — | — | |
| SEMFamily=Split-and-Merge2026.04 | 96.3 | — | — | — | — | — | |
| TableMasterTraining Dataset=PTN2023.03 | 96.04 | — | 96.16 | — | — | — | |
| RTSR2026.05 | 95.7 | — | — | — | — | 43.2 | |
| NCGMFamily=Bottom-Up2026.04 | 95.4 | — | — | — | — | — | |
| SEMTraining Dataset=PTN2023.03 | 93.7 | — | — | — | — | — | |
| GTETraining Dataset=PTN2023.03 | 93.01 | — | — | — | — | — | |
| GTEFamily=Bottom-Up2026.04 | 93 | — | — | — | — | — | |
| EDDInput=Image2019.11 | 91.2 | 85.4 | 88.3 | — | — | — | |
| EDD2021.07 | 91.2 | 85.4 | 88.3 | — | — | — | |
| OMNIPARSER V2Input Size=1,024, Decoder Len.=1,5002025.02 | 90.5 | — | 88.9 | — | — | 1.7 | |
| OmniParser2026.05 | 90.5 | — | — | — | — | 1.3 | |
| OmniparserFamily=Img-to-Seq2026.04 | 90.45 | — | 88.83 | — | — | — | |
| OmniParserInput Size=1024, Decoder Len.=1,5002025.02 | 90.4 | — | 88.8 | — | — | 1.3 | |
| OMNIPARSER V2 SPTInput Size=1,024, Decoder Len.=1,5002025.02 | 90.3 | — | 88.6 | — | — | 1.7 | |
| EDDTraining Dataset=PTN2023.03 | 89.9 | — | 88.3 | — | — | — | |
| EDDFamily=Img-to-Seq2026.04 | 89.9 | — | 88.3 | — | — | — | |
| EDDInput Size=512, Decoder Len.=1,8002025.02 | 89.9 | — | 88.3 | — | — | — | |
| EDD2026.05 | 89.9 | — | — | — | — | 1 | |
| WYGIWYSInput=Image2019.11 | 81.7 | 75.5 | 78.6 | — | — | — | |
| CamelotInput=PDF2019.11 | 80 | 66 | 73 | — | — | — | |
| TabulaInput=PDF2019.11 | 78 | 57.8 | 67.9 | — | — | — | |
| Acrobat ProInput=PDF2019.11 | 68.9 | 61.8 | 65.3 | — | — | — | |
| TraprangeInput=PDF2019.11 | 60.8 | 49.9 | 55.4 | — | — | — | |
| Acrobat ProInput=Image2019.11 | 53.8 | 53.5 | 53.7 | — | — | — | |
| PDFPlumberInput=PDF2019.11 | 44.9 | 35.9 | 40.4 | — | — | — | |
| Donut*Input Size=1,280, Decoder Len.=4,0002025.02 | 25.2 | — | 22.7 | — | — | 0.8 | |
| Acrobat ProExperimental Setup=S-A2020.10 | — | — | 53.7 | — | — | — | |
| EDDExperimental Setup=S-A, Training Dataset=PubTabNet, # Images=399K2020.10 | — | — | 88.3 | — | — | — | |
| EDD2022.10 | — | — | 88.3 | — | — | — | |
| FLAG-NetTraining Dataset=SciTSR2023.03 | — | — | 95.1 | — | — | — | |
| LGPMAModel Size (M)=1772022.10 | — | — | 94.7 | 65.74 | 96.7 | — | |
| NCGMTraining Dataset=SciTSR2023.03 | — | — | 95.4 | — | — | — | |
| SLANetInference time (ms)=766, Model Size (M)=9.22022.10 | — | — | 95.89 | 76.31 | 97.01 | — | |
| SPLERGEFamily=Split-and-Merge2026.04 | — | — | 96.27 | — | — | — | |
| TableMasterInference time (ms)=2144, Model Size (M)=2532022.10 | — | — | 96.12 | 77.9 | — | — | |
| TableMaster2026.06 | — | — | 96.32 | — | — | — | |
| TableRec-RAREInference time (ms)=779, Model Size (M)=6.82022.10 | — | — | 93.88 | 71.73 | — | — | |
| Tabstruct-NetExperimental Setup=S-A, Training Dataset=SciTSR, # Images=12K2020.10 | — | — | 90.1 | — | — | — | |
| TabStructNetTraining Dataset=SciTSR2023.03 | — | — | 90.1 | — | — | — | |
| WYGIWYSExperimental Setup=S-A, Training Dataset=PubTabNet, # Images=399K2020.10 | — | — | 78.6 | — | — | — | |
| WYGIWYSInput Size=5122025.02 | — | — | 78.6 | — | — | — |