Table Detection on ICDAR cTDaR combined archival + modern 2019 (test)
97.83mIoU@0.6DiT-L
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DiT-LBackbone=DiT-L2022.03 | 97.83 | 97.41 | 96.29 | 92.93 | 95.85 | |
| DiT-LBackbone=DiT-L, Detector Architecture=Cascade2022.03 | 97.68 | 97.26 | 97.12 | 94.74 | 96.55 | |
| DiT-BBackbone=DiT-B, Detector Architecture=Cascade2022.03 | 97.2 | 96.92 | 96.78 | 94.26 | 96.14 | |
| 1st place in cTDaR2022.03 | 96.97 | 95.99 | 95.14 | 90.22 | 94.23 | |
| MAE-BBackbone=MAE-B2022.03 | 96.86 | 96.31 | 95.05 | 91.57 | 94.66 | |
| BEIT-BBackbone=BEIT-B2022.03 | 96.82 | 96.4 | 95.41 | 92.44 | 95.03 | |
| DiT-BBackbone=DiT-B2022.03 | 96.75 | 96.19 | 95.62 | 93.36 | 95.3 | |
| ResNeXt-101-32x8dBackbone=ResNeXt-101-32x8d, Detector Architecture=Cascade2022.03 | 96.54 | 95.84 | 95.13 | 92.87 | 94.9 | |
| ResNeXt-101-32x8dBackbone=ResNeXt-101-32x8d2022.03 | 96.42 | 95.99 | 95.15 | 91.36 | 94.46 | |
| DeiT-BBackbone=DeiT-B2022.03 | 96.26 | 95.56 | 94.57 | 90.91 | 94.04 |