Physical structure recognition on ICDAR 2013
97.6PrecisionTabstruct-Net
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Tabstruct-NetTraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-B2020.10 | 97.6 | 98.5 | 98.1 | |
| DGCNNTraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-B2020.10 | 97.2 | 98.3 | 97.7 | |
| Split-PDF+HeuristicTraining Dataset=Private [10], #Images=83K, Experimental Setup=S-B2020.10 | 95.9 | 94.6 | 95.3 | |
| Split+HeuristicTraining Dataset=Private [10], #Images=83K, Experimental Setup=S-A2020.10 | 93.8 | 92.2 | 93 | |
| TableNetTraining Dataset=Marmot Extended, #Images=1K, Experimental Setup=S-B2020.10 | 92.2 | 89.9 | 91 | |
| Split-PDFTraining Dataset=Private [10], #Images=83K, Experimental Setup=S-B2020.10 | 92 | 91.3 | 91.6 | |
| Tabstruct-NetTraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-A2020.10 | 91.5 | 89.7 | 90.6 | |
| GraphTSRTraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-B2020.10 | 88.5 | 86 | 87.2 | |
| SPLERGETraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-A2020.10 | 88.3 | 87.5 | 87.9 | |
| DeepDeSRTTraining Dataset=SCITSR, #Images=12K, Experimental Setup=S-A2020.10 | 63.1 | 61.9 | 62.5 |