Physical table structure recognition on ICDAR 2019 (test)
97.5PrecisionTabStruct-Net
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| TabStruct-NetTrain Dataset=ICDAR-2019, Setup=S-B (image + ground truth boxes)2020.10 | 97.5 | 95.7 | 96.6 | |
| TabStruct-NetTrain Dataset=SciTSR+ICDAR-2019, Setup=S-B (image + ground truth boxes)2020.10 | 97.5 | 95.8 | 96.6 | |
| TabStruct-NetTrain Dataset=SciTSR, Setup=S-B (image + ground truth boxes)2020.10 | 92.4 | 89.9 | 91.1 | |
| TabStruct-NetTrain Dataset=SciTSR+ICDAR-2019, Setup=S-A (image only)2020.10 | 82.2 | 78.7 | 80.4 | |
| TabStruct-NetTrain Dataset=ICDAR-2019, Setup=S-A (image only)2020.10 | 80.3 | 76.8 | 78.5 | |
| TabStruct-NetTrain Dataset=SciTSR, Setup=S-A (image only)2020.10 | 59.5 | 57.2 | 58.3 |